Modelo Detector-Teste1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "6d9813ff-3fce-4096-ba9b-14a32eb7f008",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-13 16:20:50.530605: I tensorflow/core/util/port.cc:110] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
      "2024-04-13 16:20:50.561139: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
      "To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
      "2024-04-13 16:20:51.130180: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import json\n",
    "import itertools as it\n",
    "import time\n",
    "import subprocess\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "import tensorflow as tf\n",
    "from datetime import date, datetime, timedelta\n",
    "from tensorflow import keras\n",
    "from keras.layers import LSTM, SimpleRNN, GRU\n",
    "from sklearn.metrics import confusion_matrix,classification_report,ConfusionMatrixDisplay\n",
    "from sklearn.metrics import accuracy_score, precision_score, recall_score, roc_auc_score, average_precision_score\n",
    "from tensorflow.keras import layers, losses\n",
    "from tensorflow.keras.models import Model\n",
    "from keras.utils import plot_model\n",
    "from transformers import BertTokenizer\n",
    "from sklearn.model_selection import train_test_split\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9ba3a040-0210-4876-b2ab-2884b11f3144",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>domain</th>\n",
       "      <th>label</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>62</th>\n",
       "      <th>63</th>\n",
       "      <th>64</th>\n",
       "      <th>65</th>\n",
       "      <th>66</th>\n",
       "      <th>67</th>\n",
       "      <th>68</th>\n",
       "      <th>69</th>\n",
       "      <th>70</th>\n",
       "      <th>71</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>google.com</td>\n",
       "      <td>0</td>\n",
       "      <td>103</td>\n",
       "      <td>111</td>\n",
       "      <td>111</td>\n",
       "      <td>103</td>\n",
       "      <td>108</td>\n",
       "      <td>101</td>\n",
       "      <td>46</td>\n",
       "      <td>99</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>facebook.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>97</td>\n",
       "      <td>99</td>\n",
       "      <td>101</td>\n",
       "      <td>98</td>\n",
       "      <td>111</td>\n",
       "      <td>111</td>\n",
       "      <td>107</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>youtube.com</td>\n",
       "      <td>0</td>\n",
       "      <td>121</td>\n",
       "      <td>111</td>\n",
       "      <td>117</td>\n",
       "      <td>116</td>\n",
       "      <td>117</td>\n",
       "      <td>98</td>\n",
       "      <td>101</td>\n",
       "      <td>46</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>twitter.com</td>\n",
       "      <td>0</td>\n",
       "      <td>116</td>\n",
       "      <td>119</td>\n",
       "      <td>105</td>\n",
       "      <td>116</td>\n",
       "      <td>116</td>\n",
       "      <td>101</td>\n",
       "      <td>114</td>\n",
       "      <td>46</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>instagram.com</td>\n",
       "      <td>0</td>\n",
       "      <td>105</td>\n",
       "      <td>110</td>\n",
       "      <td>115</td>\n",
       "      <td>116</td>\n",
       "      <td>97</td>\n",
       "      <td>103</td>\n",
       "      <td>114</td>\n",
       "      <td>97</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999995</th>\n",
       "      <td>fidgit.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>105</td>\n",
       "      <td>100</td>\n",
       "      <td>103</td>\n",
       "      <td>105</td>\n",
       "      <td>116</td>\n",
       "      <td>46</td>\n",
       "      <td>99</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999996</th>\n",
       "      <td>fixwindows10connections.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>105</td>\n",
       "      <td>120</td>\n",
       "      <td>119</td>\n",
       "      <td>105</td>\n",
       "      <td>110</td>\n",
       "      <td>100</td>\n",
       "      <td>111</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999997</th>\n",
       "      <td>flyerguide.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>108</td>\n",
       "      <td>121</td>\n",
       "      <td>101</td>\n",
       "      <td>114</td>\n",
       "      <td>103</td>\n",
       "      <td>117</td>\n",
       "      <td>105</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999998</th>\n",
       "      <td>forwardpathway.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>111</td>\n",
       "      <td>114</td>\n",
       "      <td>119</td>\n",
       "      <td>97</td>\n",
       "      <td>114</td>\n",
       "      <td>100</td>\n",
       "      <td>112</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>999999</th>\n",
       "      <td>fostersmarket.com</td>\n",
       "      <td>0</td>\n",
       "      <td>102</td>\n",
       "      <td>111</td>\n",
       "      <td>115</td>\n",
       "      <td>116</td>\n",
       "      <td>101</td>\n",
       "      <td>114</td>\n",
       "      <td>115</td>\n",
       "      <td>109</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1000000 rows × 72 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                             domain  label    2    3    4    5    6    7    8  \\\n",
       "0                        google.com      0  103  111  111  103  108  101   46   \n",
       "1                      facebook.com      0  102   97   99  101   98  111  111   \n",
       "2                       youtube.com      0  121  111  117  116  117   98  101   \n",
       "3                       twitter.com      0  116  119  105  116  116  101  114   \n",
       "4                     instagram.com      0  105  110  115  116   97  103  114   \n",
       "...                             ...    ...  ...  ...  ...  ...  ...  ...  ...   \n",
       "999995                   fidgit.com      0  102  105  100  103  105  116   46   \n",
       "999996  fixwindows10connections.com      0  102  105  120  119  105  110  100   \n",
       "999997               flyerguide.com      0  102  108  121  101  114  103  117   \n",
       "999998           forwardpathway.com      0  102  111  114  119   97  114  100   \n",
       "999999            fostersmarket.com      0  102  111  115  116  101  114  115   \n",
       "\n",
       "          9  ...  62  63  64  65  66  67  68  69  70  71  \n",
       "0        99  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1       107  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "2        46  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "3        46  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "4        97  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "...     ...  ...  ..  ..  ..  ..  ..  ..  ..  ..  ..  ..  \n",
       "999995   99  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "999996  111  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "999997  105  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "999998  112  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "999999  109  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "\n",
       "[1000000 rows x 72 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "majestic = pd.read_csv('./treino/majestic_ascii.csv')\n",
    "majestic['69'] = 0\n",
    "majestic['70'] = 0\n",
    "majestic['71'] = 0\n",
    "majestic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c781d7a2-3096-43e3-b5e9-4c4943ae4958",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>domain</th>\n",
       "      <th>label</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>...</th>\n",
       "      <th>63</th>\n",
       "      <th>64</th>\n",
       "      <th>65</th>\n",
       "      <th>66</th>\n",
       "      <th>67</th>\n",
       "      <th>68</th>\n",
       "      <th>69</th>\n",
       "      <th>70</th>\n",
       "      <th>71</th>\n",
       "      <th>72</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>uzwdmrlhr.tk</td>\n",
       "      <td>1</td>\n",
       "      <td>117</td>\n",
       "      <td>122</td>\n",
       "      <td>119</td>\n",
       "      <td>100</td>\n",
       "      <td>109</td>\n",
       "      <td>114</td>\n",
       "      <td>108</td>\n",
       "      <td>104</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>uzwdmrlhr.com</td>\n",
       "      <td>1</td>\n",
       "      <td>117</td>\n",
       "      <td>122</td>\n",
       "      <td>119</td>\n",
       "      <td>100</td>\n",
       "      <td>109</td>\n",
       "      <td>114</td>\n",
       "      <td>108</td>\n",
       "      <td>104</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>uzwdmrlhr.pages.dev</td>\n",
       "      <td>1</td>\n",
       "      <td>117</td>\n",
       "      <td>122</td>\n",
       "      <td>119</td>\n",
       "      <td>100</td>\n",
       "      <td>109</td>\n",
       "      <td>114</td>\n",
       "      <td>108</td>\n",
       "      <td>104</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>rhrlwdumz.tk</td>\n",
       "      <td>1</td>\n",
       "      <td>114</td>\n",
       "      <td>104</td>\n",
       "      <td>114</td>\n",
       "      <td>108</td>\n",
       "      <td>119</td>\n",
       "      <td>100</td>\n",
       "      <td>117</td>\n",
       "      <td>109</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>rhrlwdumz.com</td>\n",
       "      <td>1</td>\n",
       "      <td>114</td>\n",
       "      <td>104</td>\n",
       "      <td>114</td>\n",
       "      <td>108</td>\n",
       "      <td>119</td>\n",
       "      <td>100</td>\n",
       "      <td>117</td>\n",
       "      <td>109</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1352254</th>\n",
       "      <td>kwgjjpbcaaanbivklyxx.com</td>\n",
       "      <td>1</td>\n",
       "      <td>107</td>\n",
       "      <td>119</td>\n",
       "      <td>103</td>\n",
       "      <td>106</td>\n",
       "      <td>106</td>\n",
       "      <td>112</td>\n",
       "      <td>98</td>\n",
       "      <td>99</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1352255</th>\n",
       "      <td>vxijfvgvihjnfvofxmyc.com</td>\n",
       "      <td>1</td>\n",
       "      <td>118</td>\n",
       "      <td>120</td>\n",
       "      <td>105</td>\n",
       "      <td>106</td>\n",
       "      <td>102</td>\n",
       "      <td>118</td>\n",
       "      <td>103</td>\n",
       "      <td>118</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1352256</th>\n",
       "      <td>cepswmbgtknhtkhlsecg.com</td>\n",
       "      <td>1</td>\n",
       "      <td>99</td>\n",
       "      <td>101</td>\n",
       "      <td>112</td>\n",
       "      <td>115</td>\n",
       "      <td>119</td>\n",
       "      <td>109</td>\n",
       "      <td>98</td>\n",
       "      <td>103</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1352257</th>\n",
       "      <td>tifjhjtouqmdlbkfnuow.com</td>\n",
       "      <td>1</td>\n",
       "      <td>116</td>\n",
       "      <td>105</td>\n",
       "      <td>102</td>\n",
       "      <td>106</td>\n",
       "      <td>104</td>\n",
       "      <td>106</td>\n",
       "      <td>116</td>\n",
       "      <td>111</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1352258</th>\n",
       "      <td>gjrykbribvusoskfnuow.com</td>\n",
       "      <td>1</td>\n",
       "      <td>103</td>\n",
       "      <td>106</td>\n",
       "      <td>114</td>\n",
       "      <td>121</td>\n",
       "      <td>107</td>\n",
       "      <td>98</td>\n",
       "      <td>114</td>\n",
       "      <td>105</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1352259 rows × 72 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                           domain  label    3    4    5    6    7    8    9  \\\n",
       "0                    uzwdmrlhr.tk      1  117  122  119  100  109  114  108   \n",
       "1                   uzwdmrlhr.com      1  117  122  119  100  109  114  108   \n",
       "2             uzwdmrlhr.pages.dev      1  117  122  119  100  109  114  108   \n",
       "3                    rhrlwdumz.tk      1  114  104  114  108  119  100  117   \n",
       "4                   rhrlwdumz.com      1  114  104  114  108  119  100  117   \n",
       "...                           ...    ...  ...  ...  ...  ...  ...  ...  ...   \n",
       "1352254  kwgjjpbcaaanbivklyxx.com      1  107  119  103  106  106  112   98   \n",
       "1352255  vxijfvgvihjnfvofxmyc.com      1  118  120  105  106  102  118  103   \n",
       "1352256  cepswmbgtknhtkhlsecg.com      1   99  101  112  115  119  109   98   \n",
       "1352257  tifjhjtouqmdlbkfnuow.com      1  116  105  102  106  104  106  116   \n",
       "1352258  gjrykbribvusoskfnuow.com      1  103  106  114  121  107   98  114   \n",
       "\n",
       "          10  ...  63  64  65  66  67  68  69  70  71  72  \n",
       "0        104  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1        104  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "2        104  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "3        109  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "4        109  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "...      ...  ...  ..  ..  ..  ..  ..  ..  ..  ..  ..  ..  \n",
       "1352254   99  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1352255  118  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1352256  103  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1352257  111  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "1352258  105  ...   0   0   0   0   0   0   0   0   0   0  \n",
       "\n",
       "[1352259 rows x 72 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dga = pd.read_csv('./treino/df_train.csv')\n",
    "dga = dga.drop('family', axis = 1)\n",
    "dga"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5676e746-4851-453d-8369-c5e57648d0d8",
   "metadata": {},
   "outputs": [],
   "source": [
    "dga.columns = majestic.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "e163dad2-01d6-4a82-8adf-4c0a32760000",
   "metadata": {},
   "outputs": [],
   "source": [
    "dga = dga.loc[~dga['domain'].isin(majestic['domain'])].reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9cb322d2-17c3-4a61-9254-38e7870d61fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "dga = pd.concat([majestic,dga], axis = 0, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3873000b-8f11-4818-bbf1-143b061b01a4",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train,df_test = train_test_split(dga, test_size=0.3, random_state=1, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5a7034d5-83a1-4a07-845c-59afb8afcee9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>domain</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>...</th>\n",
       "      <th>62</th>\n",
       "      <th>63</th>\n",
       "      <th>64</th>\n",
       "      <th>65</th>\n",
       "      <th>66</th>\n",
       "      <th>67</th>\n",
       "      <th>68</th>\n",
       "      <th>69</th>\n",
       "      <th>70</th>\n",
       "      <th>71</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>label</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>...</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "      <td>700179</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>...</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "      <td>946365</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 71 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       domain       2       3       4       5       6       7       8       9  \\\n",
       "label                                                                           \n",
       "0      700179  700179  700179  700179  700179  700179  700179  700179  700179   \n",
       "1      946365  946365  946365  946365  946365  946365  946365  946365  946365   \n",
       "\n",
       "           10  ...      62      63      64      65      66      67      68  \\\n",
       "label          ...                                                           \n",
       "0      700179  ...  700179  700179  700179  700179  700179  700179  700179   \n",
       "1      946365  ...  946365  946365  946365  946365  946365  946365  946365   \n",
       "\n",
       "           69      70      71  \n",
       "label                          \n",
       "0      700179  700179  700179  \n",
       "1      946365  946365  946365  \n",
       "\n",
       "[2 rows x 71 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train.groupby('label').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "042fd7b6-7383-45bf-b6d1-b6ab28716bba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>domain</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>10</th>\n",
       "      <th>...</th>\n",
       "      <th>62</th>\n",
       "      <th>63</th>\n",
       "      <th>64</th>\n",
       "      <th>65</th>\n",
       "      <th>66</th>\n",
       "      <th>67</th>\n",
       "      <th>68</th>\n",
       "      <th>69</th>\n",
       "      <th>70</th>\n",
       "      <th>71</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>label</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>...</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "      <td>299821</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>...</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "      <td>405842</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 71 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       domain       2       3       4       5       6       7       8       9  \\\n",
       "label                                                                           \n",
       "0      299821  299821  299821  299821  299821  299821  299821  299821  299821   \n",
       "1      405842  405842  405842  405842  405842  405842  405842  405842  405842   \n",
       "\n",
       "           10  ...      62      63      64      65      66      67      68  \\\n",
       "label          ...                                                           \n",
       "0      299821  ...  299821  299821  299821  299821  299821  299821  299821   \n",
       "1      405842  ...  405842  405842  405842  405842  405842  405842  405842   \n",
       "\n",
       "           69      70      71  \n",
       "label                          \n",
       "0      299821  299821  299821  \n",
       "1      405842  405842  405842  \n",
       "\n",
       "[2 rows x 71 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test.groupby('label').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "94f6b43c-2654-4f8a-a87f-89fa44381693",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = df_train.drop(['domain','label'], axis = 1)\n",
    "y_train = df_train['label']\n",
    "\n",
    "X_test = df_test.drop(['domain','label'], axis = 1)\n",
    "y_test = df_test['label']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "654d8a89-38de-49b0-8624-f1657e8d6e49",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.to_numpy()\n",
    "y_train = y_train.to_numpy()\n",
    "\n",
    "X_test = X_test.to_numpy()\n",
    "y_test = y_test.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "79c31399-2610-462f-afc6-1b9a5567765a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((1646544, 70), (705663, 70))"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape, X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ea436f37-3304-4f6f-a606-7306acf8818e",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 14:30:53.538687: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.582817: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.582966: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.584732: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.584857: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.584935: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.854210: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.854345: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.854424: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-10 14:30:53.854497: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1639] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 1857 MB memory:  -> device: 0, name: NVIDIA GeForce RTX 3050 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6\n"
     ]
    }
   ],
   "source": [
    "#Carregando hyperparâmetros de regularizadores, otimizador e função de ativação\n",
    "l1l2 = keras.regularizers.l1_l2(l1=0.00001, l2=0.0001)\n",
    "opt = keras.optimizers.Adam(learning_rate=0.0001)\n",
    "fx = \"relu\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b6113b02-7c42-4dcf-806f-84a90c055364",
   "metadata": {},
   "source": [
    "## Modelo 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "c4159eaf-763d-4461-831d-1363f7bffcb1",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 70, 60)            15360     \n",
      "                                                                 \n",
      " conv1d (Conv1D)             (None, 70, 500)           210500    \n",
      "                                                                 \n",
      " re_lu (ReLU)                (None, 70, 500)           0         \n",
      "                                                                 \n",
      " conv1d_1 (Conv1D)           (None, 70, 250)           750250    \n",
      "                                                                 \n",
      " re_lu_1 (ReLU)              (None, 70, 250)           0         \n",
      "                                                                 \n",
      " max_pooling1d (MaxPooling1  (None, 70, 250)           0         \n",
      " D)                                                              \n",
      "                                                                 \n",
      " conv1d_2 (Conv1D)           (None, 70, 120)           150120    \n",
      "                                                                 \n",
      " re_lu_2 (ReLU)              (None, 70, 120)           0         \n",
      "                                                                 \n",
      " conv1d_3 (Conv1D)           (None, 70, 60)            36060     \n",
      "                                                                 \n",
      " re_lu_3 (ReLU)              (None, 70, 60)            0         \n",
      "                                                                 \n",
      " max_pooling1d_1 (MaxPoolin  (None, 70, 60)            0         \n",
      " g1D)                                                            \n",
      "                                                                 \n",
      " conv1d_4 (Conv1D)           (None, 70, 30)            7230      \n",
      "                                                                 \n",
      " re_lu_4 (ReLU)              (None, 70, 30)            0         \n",
      "                                                                 \n",
      " conv1d_5 (Conv1D)           (None, 70, 30)            2730      \n",
      "                                                                 \n",
      " re_lu_5 (ReLU)              (None, 70, 30)            0         \n",
      "                                                                 \n",
      " max_pooling1d_2 (MaxPoolin  (None, 70, 30)            0         \n",
      " g1D)                                                            \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 2100)              0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 15)                31515     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 1)                 16        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1203781 (4.59 MB)\n",
      "Trainable params: 1203781 (4.59 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Estrutura do modelo convolucional profundo\n",
    "model_1 = tf.keras.Sequential([\n",
    "      \n",
    "        layers.Embedding(input_dim=256,output_dim=60,input_length=70),\n",
    "    \n",
    "        layers.Conv1D(500, 7, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "            \n",
    "        layers.Conv1D(250, 6, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=6, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(120, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(60, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=4, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(30, 4, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(30, 3, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=3, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Flatten(),\n",
    "        layers.Dense(15, activation=fx, activity_regularizer=l1l2),\n",
    "        layers.Dense(1, activation='sigmoid')])\n",
    "\n",
    "model_1.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_1.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "d9f0a8ad-d1a5-4bce-b4f8-231f9ea84a9c",
   "metadata": {},
   "outputs": [],
   "source": [
    "best11 = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model11',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "459b7eea-c733-41de-bf90-224511faae52",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Ajustando uso de memoria RAM e GPU\n",
    "physical_devices = tf.config.list_physical_devices('GPU')\n",
    "try:\n",
    "  tf.config.experimental.set_memory_growth(physical_devices[0], True)\n",
    "except:\n",
    "  pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "749df9b4-1f90-46f1-9dc9-541e264473e2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 08:19:20.690771: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-10 08:19:21.342764: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "2024-04-10 08:19:21.344925: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x5c1a15d95810 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-10 08:19:21.344937: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-10 08:19:21.348546: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-10 08:19:21.432124: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n",
      "2024-04-10 08:19:21.790514: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.50GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 08:19:22.095325: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.25GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 08:19:22.377733: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 4.48GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1078 - accuracy: 0.9607"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 08:24:05.624739: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 08:24:05.630357: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 298s 40ms/step - loss: 0.1078 - accuracy: 0.9607 - val_loss: 0.0709 - val_accuracy: 0.9757\n",
      "Epoch 2/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.0629 - accuracy: 0.9785INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 289s 39ms/step - loss: 0.0629 - accuracy: 0.9785 - val_loss: 0.0629 - val_accuracy: 0.9782\n",
      "Epoch 3/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.0508 - accuracy: 0.9828INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 288s 39ms/step - loss: 0.0508 - accuracy: 0.9828 - val_loss: 0.0611 - val_accuracy: 0.9784\n",
      "Epoch 4/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0426 - accuracy: 0.9856INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 288s 39ms/step - loss: 0.0426 - accuracy: 0.9856 - val_loss: 0.0612 - val_accuracy: 0.9788\n",
      "Epoch 5/10\n",
      "7410/7410 [==============================] - 287s 39ms/step - loss: 0.0361 - accuracy: 0.9880 - val_loss: 0.0645 - val_accuracy: 0.9785\n",
      "Epoch 6/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0306 - accuracy: 0.9900INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model11/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 288s 39ms/step - loss: 0.0306 - accuracy: 0.9900 - val_loss: 0.0626 - val_accuracy: 0.9796\n",
      "Epoch 7/10\n",
      "7410/7410 [==============================] - 287s 39ms/step - loss: 0.0261 - accuracy: 0.9915 - val_loss: 0.0655 - val_accuracy: 0.9788\n",
      "Epoch 8/10\n",
      "7410/7410 [==============================] - 287s 39ms/step - loss: 0.0223 - accuracy: 0.9929 - val_loss: 0.0725 - val_accuracy: 0.9789\n",
      "Epoch 9/10\n",
      "7410/7410 [==============================] - 288s 39ms/step - loss: 0.0191 - accuracy: 0.9940 - val_loss: 0.0733 - val_accuracy: 0.9785\n",
      "Epoch 10/10\n",
      "7410/7410 [==============================] - 290s 39ms/step - loss: 0.0165 - accuracy: 0.9950 - val_loss: 0.0754 - val_accuracy: 0.9792\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x7c00dfdc4250>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_1.fit(X_train1, y_train1,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best11],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "e9ebb56a-d2f1-444a-8f7a-135ae9f543b6",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_1 = tf.keras.models.load_model('best_model11')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "89897311-76b1-4c02-806c-c473cffa620c",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 09:16:25.961880: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   73/22052 [..............................] - ETA: 46s  "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 09:16:26.405279: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 47s 2ms/step\n",
      "Matriz de Confusão\n",
      "[[292883   6938]\n",
      " [  7183 398659]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9761    0.9769    0.9765    299821\n",
      "           1     0.9829    0.9823    0.9826    405842\n",
      "\n",
      "    accuracy                         0.9800    705663\n",
      "   macro avg     0.9795    0.9796    0.9795    705663\n",
      "weighted avg     0.9800    0.9800    0.9800    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_1.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d7ed3a1-68aa-4a9d-85db-c5baf149db04",
   "metadata": {},
   "source": [
    "## Modelo 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "b116015f-c131-4053-b039-b99c34959aad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 70, 60)            15360     \n",
      "                                                                 \n",
      " conv1d (Conv1D)             (None, 70, 500)           210500    \n",
      "                                                                 \n",
      " conv1d_1 (Conv1D)           (None, 70, 250)           750250    \n",
      "                                                                 \n",
      " conv1d_2 (Conv1D)           (None, 70, 120)           150120    \n",
      "                                                                 \n",
      " conv1d_3 (Conv1D)           (None, 70, 60)            36060     \n",
      "                                                                 \n",
      " conv1d_4 (Conv1D)           (None, 70, 30)            7230      \n",
      "                                                                 \n",
      " conv1d_5 (Conv1D)           (None, 70, 30)            2730      \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 2100)              0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 15)                31515     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 1)                 16        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 1203781 (4.59 MB)\n",
      "Trainable params: 1203781 (4.59 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Estrutura do modelo convolucional profundo\n",
    "model_2 = tf.keras.Sequential([\n",
    "      \n",
    "        layers.Embedding(input_dim=256,output_dim=60,input_length=70),\n",
    "    \n",
    "        layers.Conv1D(500, 7, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "            \n",
    "        layers.Conv1D(250, 6, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "        #layers.MaxPooling1D(pool_size=6, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(120, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(60, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "        #layers.MaxPooling1D(pool_size=4, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(30, 4, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(30, 3, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        #layers.ReLU(),\n",
    "        #layers.MaxPooling1D(pool_size=3, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Flatten(),\n",
    "        layers.Dense(15, activation=fx, activity_regularizer=l1l2),\n",
    "        layers.Dense(1, activation='sigmoid')])\n",
    "\n",
    "model_2.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_2.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "16442558-e7c2-4e15-8b3d-e7bb085683ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "best21 = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model21',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "30ceaf8a-7fee-4220-95ea-4da977bc9593",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 09:24:06.688514: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-10 09:24:07.328327: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "2024-04-10 09:24:07.330776: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x7959b0002a40 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-10 09:24:07.330792: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-10 09:24:07.334570: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-10 09:24:07.414341: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n",
      "2024-04-10 09:24:07.754168: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.50GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 09:24:08.048779: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.25GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 09:24:08.326663: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 4.48GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1162 - accuracy: 0.9579"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 09:28:21.797507: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 09:28:21.803091: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 266s 35ms/step - loss: 0.1162 - accuracy: 0.9579 - val_loss: 0.0751 - val_accuracy: 0.9747\n",
      "Epoch 2/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0679 - accuracy: 0.9769INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 257s 35ms/step - loss: 0.0679 - accuracy: 0.9769 - val_loss: 0.0682 - val_accuracy: 0.9759\n",
      "Epoch 3/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0546 - accuracy: 0.9815INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.0546 - accuracy: 0.9815 - val_loss: 0.0607 - val_accuracy: 0.9795\n",
      "Epoch 4/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0452 - accuracy: 0.9847INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model21/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.0452 - accuracy: 0.9847 - val_loss: 0.0587 - val_accuracy: 0.9798\n",
      "Epoch 5/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.0373 - accuracy: 0.9876 - val_loss: 0.0614 - val_accuracy: 0.9793\n",
      "Epoch 6/10\n",
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.0308 - accuracy: 0.9900 - val_loss: 0.0636 - val_accuracy: 0.9790\n",
      "Epoch 7/10\n",
      "7410/7410 [==============================] - 260s 35ms/step - loss: 0.0254 - accuracy: 0.9919 - val_loss: 0.0745 - val_accuracy: 0.9782\n",
      "Epoch 8/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.0212 - accuracy: 0.9934 - val_loss: 0.0711 - val_accuracy: 0.9788\n",
      "Epoch 9/10\n",
      "7410/7410 [==============================] - 257s 35ms/step - loss: 0.0181 - accuracy: 0.9945 - val_loss: 0.0766 - val_accuracy: 0.9780\n",
      "Epoch 10/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.0155 - accuracy: 0.9953 - val_loss: 0.0766 - val_accuracy: 0.9782\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x795bf75c3f10>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_2.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best21],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "ef258453-e322-4928-a0ce-5bbf18e5081f",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_2 = tf.keras.models.load_model('best_model21')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "180c62f4-e8ee-4426-a029-8baecf34adb6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 43s 2ms/step\n",
      "Matriz de Confusão\n",
      "[[292879   6942]\n",
      " [  7239 398603]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9759    0.9768    0.9764    299821\n",
      "           1     0.9829    0.9822    0.9825    405842\n",
      "\n",
      "    accuracy                         0.9799    705663\n",
      "   macro avg     0.9794    0.9795    0.9794    705663\n",
      "weighted avg     0.9799    0.9799    0.9799    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_2.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e7333b61-d8f7-464a-ba7b-b7059ef54c55",
   "metadata": {},
   "source": [
    "## Modelo 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "3315a3d5-8f6d-407b-9bb0-06742c482aba",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.reshape(X_train.shape[0],X_train.shape[1],1)\n",
    "X_test = X_test.reshape(X_test.shape[0],X_test.shape[1],1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "924beb92-f818-44c2-8317-6d4a26f866a7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " conv1d (Conv1D)             (None, 70, 500)           4000      \n",
      "                                                                 \n",
      " re_lu (ReLU)                (None, 70, 500)           0         \n",
      "                                                                 \n",
      " conv1d_1 (Conv1D)           (None, 70, 250)           750250    \n",
      "                                                                 \n",
      " re_lu_1 (ReLU)              (None, 70, 250)           0         \n",
      "                                                                 \n",
      " max_pooling1d (MaxPooling1  (None, 70, 250)           0         \n",
      " D)                                                              \n",
      "                                                                 \n",
      " conv1d_2 (Conv1D)           (None, 70, 120)           150120    \n",
      "                                                                 \n",
      " re_lu_2 (ReLU)              (None, 70, 120)           0         \n",
      "                                                                 \n",
      " conv1d_3 (Conv1D)           (None, 70, 60)            36060     \n",
      "                                                                 \n",
      " re_lu_3 (ReLU)              (None, 70, 60)            0         \n",
      "                                                                 \n",
      " max_pooling1d_1 (MaxPoolin  (None, 70, 60)            0         \n",
      " g1D)                                                            \n",
      "                                                                 \n",
      " conv1d_4 (Conv1D)           (None, 70, 30)            7230      \n",
      "                                                                 \n",
      " re_lu_4 (ReLU)              (None, 70, 30)            0         \n",
      "                                                                 \n",
      " conv1d_5 (Conv1D)           (None, 70, 30)            2730      \n",
      "                                                                 \n",
      " re_lu_5 (ReLU)              (None, 70, 30)            0         \n",
      "                                                                 \n",
      " max_pooling1d_2 (MaxPoolin  (None, 70, 30)            0         \n",
      " g1D)                                                            \n",
      "                                                                 \n",
      " flatten (Flatten)           (None, 2100)              0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 15)                31515     \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 1)                 16        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 981921 (3.75 MB)\n",
      "Trainable params: 981921 (3.75 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#Estrutura do modelo convolucional profundo\n",
    "model_3 = tf.keras.Sequential([\n",
    "      \n",
    "        layers.Input(shape=(X_train.shape[1],X_train.shape[2])),\n",
    "    \n",
    "        layers.Conv1D(500, 7, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "            \n",
    "        layers.Conv1D(250, 6, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=6, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(120, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(60, 5, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=4, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Conv1D(30, 4, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "    \n",
    "        layers.Conv1D(30, 3, activation=fx, activity_regularizer=l1l2, padding='same'),\n",
    "        layers.ReLU(),\n",
    "        layers.MaxPooling1D(pool_size=3, strides=1, padding='same'),\n",
    "    \n",
    "        layers.Flatten(),\n",
    "        layers.Dense(15, activation=fx, activity_regularizer=l1l2),\n",
    "        layers.Dense(1, activation='sigmoid')])\n",
    "\n",
    "model_3.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_3.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "9c6e613d-c75f-4808-8036-26ef1d47c9a9",
   "metadata": {},
   "outputs": [],
   "source": [
    "best31 = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model31',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "63c24956-4fc4-4a3f-8514-f14d5182b59f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 10:16:07.678209: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-10 10:16:08.252077: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "2024-04-10 10:16:08.254196: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x5d4cbbf73700 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-10 10:16:08.254209: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-10 10:16:08.257850: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-10 10:16:08.337535: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n",
      "2024-04-10 10:16:08.678287: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.50GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 10:16:08.972095: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 2.25GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 10:16:09.253216: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 4.48GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.5356 - accuracy: 0.8476"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 10:20:19.781725: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 5.53GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 10:20:19.781767: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 5.53GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - ETA: 0s - loss: 0.5356 - accuracy: 0.8476"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 10:20:20.788327: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 5.55GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 10:20:21.490875: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n",
      "2024-04-10 10:20:21.496431: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 1.73GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 265s 35ms/step - loss: 0.5356 - accuracy: 0.8476 - val_loss: 0.3063 - val_accuracy: 0.8760\n",
      "Epoch 2/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2947 - accuracy: 0.8798INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.2947 - accuracy: 0.8798 - val_loss: 0.2707 - val_accuracy: 0.8915\n",
      "Epoch 3/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.2709 - accuracy: 0.8904 - val_loss: 0.2987 - val_accuracy: 0.8770\n",
      "Epoch 4/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2575 - accuracy: 0.8962INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.2575 - accuracy: 0.8962 - val_loss: 0.2475 - val_accuracy: 0.9011\n",
      "Epoch 5/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.2471 - accuracy: 0.9012 - val_loss: 0.2646 - val_accuracy: 0.8928\n",
      "Epoch 6/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.2391 - accuracy: 0.9048 - val_loss: 0.2688 - val_accuracy: 0.8897\n",
      "Epoch 7/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2325 - accuracy: 0.9078INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.2325 - accuracy: 0.9078 - val_loss: 0.2321 - val_accuracy: 0.9082\n",
      "Epoch 8/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2278 - accuracy: 0.9099INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.2278 - accuracy: 0.9099 - val_loss: 0.2211 - val_accuracy: 0.9143\n",
      "Epoch 9/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2229 - accuracy: 0.9120INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model31/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 259s 35ms/step - loss: 0.2229 - accuracy: 0.9120 - val_loss: 0.2161 - val_accuracy: 0.9158\n",
      "Epoch 10/10\n",
      "7410/7410 [==============================] - 258s 35ms/step - loss: 0.2186 - accuracy: 0.9140 - val_loss: 0.2197 - val_accuracy: 0.9135\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x7f0da3875b40>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_3.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best31],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "d61b03ea-bba2-4760-a18f-c8e975e92cbe",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_3 = tf.keras.models.load_model('best_model31')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "1e3eb089-5b50-4dd7-9085-3c0745cfb649",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 43s 2ms/step\n",
      "Matriz de Confusão\n",
      "[[265703  34118]\n",
      " [ 25522 380320]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9124    0.8862    0.8991    299821\n",
      "           1     0.9177    0.9371    0.9273    405842\n",
      "\n",
      "    accuracy                         0.9155    705663\n",
      "   macro avg     0.9150    0.9117    0.9132    705663\n",
      "weighted avg     0.9154    0.9155    0.9153    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_3.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7840f8b1-4116-4c0a-a51a-3c736483774c",
   "metadata": {},
   "source": [
    "## Bilbo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "7c63a9fd-c9de-426b-81a9-d39ed6005ef9",
   "metadata": {},
   "outputs": [],
   "source": [
    "MAX_INDEX = 70 # Max number of valid characters in a domain\n",
    "MAX_STRING_LENGTH = 63\n",
    "EMBEDDING_DIMENSION = 128\n",
    "NUM_CONV_FILTERS = 60\n",
    "\n",
    "#Carregando hyperparâmetros de regularizadores, otimizador e função de ativação\n",
    "l1l2 = keras.regularizers.l1_l2(l1=0.00001, l2=0.0001)\n",
    "opt = keras.optimizers.legacy.Adam(learning_rate=0.0001)\n",
    "fx = \"relu\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "5861a019-5ef1-44a9-befd-83270bd68368",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                Output Shape                 Param #   Connected to                  \n",
      "==================================================================================================\n",
      " input (InputLayer)          [(None, 70)]                 0         []                            \n",
      "                                                                                                  \n",
      " embeddingCNN (Embedding)    (None, 70, 128)              8960      ['input[0][0]']               \n",
      "                                                                                                  \n",
      " conv2 (Conv1D)              (None, 69, 60)               15420     ['embeddingCNN[0][0]']        \n",
      "                                                                                                  \n",
      " conv3 (Conv1D)              (None, 68, 60)               23100     ['embeddingCNN[0][0]']        \n",
      "                                                                                                  \n",
      " conv4 (Conv1D)              (None, 67, 60)               30780     ['embeddingCNN[0][0]']        \n",
      "                                                                                                  \n",
      " conv5 (Conv1D)              (None, 66, 60)               38460     ['embeddingCNN[0][0]']        \n",
      "                                                                                                  \n",
      " conv6 (Conv1D)              (None, 65, 60)               46140     ['embeddingCNN[0][0]']        \n",
      "                                                                                                  \n",
      " pool2 (GlobalMaxPooling1D)  (None, 60)                   0         ['conv2[0][0]']               \n",
      "                                                                                                  \n",
      " pool3 (GlobalMaxPooling1D)  (None, 60)                   0         ['conv3[0][0]']               \n",
      "                                                                                                  \n",
      " pool4 (GlobalMaxPooling1D)  (None, 60)                   0         ['conv4[0][0]']               \n",
      "                                                                                                  \n",
      " pool5 (GlobalMaxPooling1D)  (None, 60)                   0         ['conv5[0][0]']               \n",
      "                                                                                                  \n",
      " pool6 (GlobalMaxPooling1D)  (None, 60)                   0         ['conv6[0][0]']               \n",
      "                                                                                                  \n",
      " concatenate_2 (Concatenate  (None, 300)                  0         ['pool2[0][0]',               \n",
      " )                                                                   'pool3[0][0]',               \n",
      "                                                                     'pool4[0][0]',               \n",
      "                                                                     'pool5[0][0]',               \n",
      "                                                                     'pool6[0][0]']               \n",
      "                                                                                                  \n",
      " dropoutcnnmid (Dropout)     (None, 300)                  0         ['concatenate_2[0][0]']       \n",
      "                                                                                                  \n",
      " embeddingLSTM (Embedding)   (None, 70, 20)               5120      ['input[0][0]']               \n",
      "                                                                                                  \n",
      " densecnn (Dense)            (None, 60)                   18060     ['dropoutcnnmid[0][0]']       \n",
      "                                                                                                  \n",
      " lstm (LSTM)                 (None, 256)                  283648    ['embeddingLSTM[0][0]']       \n",
      "                                                                                                  \n",
      " dropoutcnn (Dropout)        (None, 60)                   0         ['densecnn[0][0]']            \n",
      "                                                                                                  \n",
      " dropoutlstm (Dropout)       (None, 256)                  0         ['lstm[0][0]']                \n",
      "                                                                                                  \n",
      " concatenate_3 (Concatenate  (None, 316)                  0         ['dropoutcnn[0][0]',          \n",
      " )                                                                   'dropoutlstm[0][0]']         \n",
      "                                                                                                  \n",
      " dropoutsemifinal (Dropout)  (None, 316)                  0         ['concatenate_3[0][0]']       \n",
      "                                                                                                  \n",
      " extradense (Dense)          (None, 100)                  31700     ['dropoutsemifinal[0][0]']    \n",
      "                                                                                                  \n",
      " dropoutfinal (Dropout)      (None, 100)                  0         ['extradense[0][0]']          \n",
      "                                                                                                  \n",
      " output (Dense)              (None, 1)                    101       ['dropoutfinal[0][0]']        \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 501489 (1.91 MB)\n",
      "Trainable params: 501489 (1.91 MB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input = layers.Input((70,), dtype='int32', name=\"input\")\n",
    "\n",
    "########################\n",
    "#          CNN         #\n",
    "########################\n",
    "\n",
    "embeddingCNN = layers.Embedding(\n",
    "                        output_dim=EMBEDDING_DIMENSION, \n",
    "                        input_dim=MAX_INDEX,\n",
    "                        input_length=MAX_STRING_LENGTH, \n",
    "                        name=\"embeddingCNN\")(input)\n",
    "\n",
    "# Parallel Convolutional Layer\n",
    "\n",
    "conv2 = layers.Conv1D(NUM_CONV_FILTERS, 2, name=\"conv2\") (embeddingCNN)\n",
    "\n",
    "conv3 = layers.Conv1D(NUM_CONV_FILTERS, 3, name=\"conv3\") (embeddingCNN)\n",
    "\n",
    "conv4 = layers.Conv1D(NUM_CONV_FILTERS, 4, name=\"conv4\") (embeddingCNN)\n",
    "\n",
    "conv5 = layers.Conv1D(NUM_CONV_FILTERS, 5, name=\"conv5\") (embeddingCNN)\n",
    "\n",
    "conv6 = layers.Conv1D(NUM_CONV_FILTERS, 6, name=\"conv6\") (embeddingCNN)\n",
    "\n",
    "# Global max pooling\n",
    "\n",
    "pool2 = layers.GlobalMaxPool1D(name=\"pool2\") (conv2)\n",
    "\n",
    "pool3 = layers.GlobalMaxPool1D(name=\"pool3\") (conv3)\n",
    "\n",
    "pool4 = layers.GlobalMaxPool1D(name=\"pool4\") (conv4)\n",
    "\n",
    "pool5 = layers.GlobalMaxPool1D(name=\"pool5\") (conv5)\n",
    "\n",
    "pool6 = layers.GlobalMaxPool1D(name=\"pool6\") (conv6)\n",
    "\n",
    "\n",
    "concatcnn = layers.Concatenate(axis=1)([pool2,pool3, pool4, pool5, pool6])\n",
    "\n",
    "\n",
    "dropoutcnnmid = layers.Dropout(0.5, name=\"dropoutcnnmid\")(concatcnn)\n",
    "\n",
    "densecnn = layers.Dense(NUM_CONV_FILTERS, activation=\"relu\",\n",
    "                        name=\"densecnn\")(dropoutcnnmid)\n",
    "\n",
    "dropoutcnn = layers.Dropout(0.5, name=\"dropoutcnn\") (densecnn)\n",
    "\n",
    "########################\n",
    "#         LSTM         #\n",
    "########################\n",
    "\n",
    "embeddingLSTM = layers.Embedding(output_dim=20, \n",
    "                             input_dim=256,\n",
    "                             input_length=MAX_STRING_LENGTH, \n",
    "                             name=\"embeddingLSTM\") (input)\n",
    "\n",
    "\n",
    "lstm = layers.LSTM(256, name=\"lstm\")(embeddingLSTM)\n",
    "\n",
    "dropoutlstm = layers.Dropout(0.5, name=\"dropoutlstm\") (lstm)\n",
    "\n",
    "########################\n",
    "#    Combine - ANN     #\n",
    "########################\n",
    "\n",
    "concat = layers.Concatenate(axis=-1)([dropoutcnn, dropoutlstm])\n",
    "\n",
    "dropoutsemifinal = layers.Dropout(0.5, name=\"dropoutsemifinal\") (concat)\n",
    "\n",
    "extradense = layers.Dense(100, activation='relu', \n",
    "                        name=\"extradense\") (dropoutsemifinal)\n",
    "\n",
    "dropoutfinal = layers.Dropout(0.5, name=\"dropoutfinal\") (extradense)\n",
    "\n",
    "output = layers.Dense(1, activation='sigmoid', name=\"output\") (dropoutfinal)\n",
    "\n",
    "model_bilbo = Model(input,output)\n",
    "\n",
    "model_bilbo.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_bilbo.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "99488a75-f5b2-446b-97df-e00c23eb38ff",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "plot_model(model_bilbo, to_file='model_plot.png', show_shapes=True, show_layer_names=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "1a811fbb-950c-4063-9b10-355e241fa82e",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_bilbo1 = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model_bilbo1',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "bde8d1a3-446d-4d1f-8bc7-678c552b2747",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 11:03:01.379831: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 4.68GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "   4/7410 [..............................] - ETA: 2:15 - loss: 0.6922 - accuracy: 0.5050   "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 11:03:01.852912: W tensorflow/tsl/framework/bfc_allocator.cc:296] Allocator (GPU_0_bfc) ran out of memory trying to allocate 4.68GiB with freed_by_count=0. The caller indicates that this is not a failure, but this may mean that there could be performance gains if more memory were available.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - ETA: 0s - loss: 0.3149 - accuracy: 0.8530INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 145s 19ms/step - loss: 0.3149 - accuracy: 0.8530 - val_loss: 0.2431 - val_accuracy: 0.9009\n",
      "Epoch 2/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.2289 - accuracy: 0.9080INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.2289 - accuracy: 0.9080 - val_loss: 0.2044 - val_accuracy: 0.9176\n",
      "Epoch 3/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.2000 - accuracy: 0.9214INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.2000 - accuracy: 0.9214 - val_loss: 0.1776 - val_accuracy: 0.9310\n",
      "Epoch 4/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1792 - accuracy: 0.9310INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 143s 19ms/step - loss: 0.1792 - accuracy: 0.9310 - val_loss: 0.1619 - val_accuracy: 0.9366\n",
      "Epoch 5/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1639 - accuracy: 0.9373INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.1639 - accuracy: 0.9373 - val_loss: 0.1522 - val_accuracy: 0.9402\n",
      "Epoch 6/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1519 - accuracy: 0.9419INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.1519 - accuracy: 0.9419 - val_loss: 0.1377 - val_accuracy: 0.9471\n",
      "Epoch 7/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1433 - accuracy: 0.9454INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.1433 - accuracy: 0.9454 - val_loss: 0.1313 - val_accuracy: 0.9496\n",
      "Epoch 8/10\n",
      "7410/7410 [==============================] - 140s 19ms/step - loss: 0.1356 - accuracy: 0.9487 - val_loss: 0.1296 - val_accuracy: 0.9477\n",
      "Epoch 9/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1286 - accuracy: 0.9515INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.1286 - accuracy: 0.9515 - val_loss: 0.1224 - val_accuracy: 0.9528\n",
      "Epoch 10/10\n",
      "7407/7410 [============================>.] - ETA: 0s - loss: 0.1224 - accuracy: 0.9541INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bilbo1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 142s 19ms/step - loss: 0.1224 - accuracy: 0.9541 - val_loss: 0.1186 - val_accuracy: 0.9552\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x7f0b9014ff40>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_bilbo.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best_bilbo1],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "3aece5a0-7b57-4fff-98b0-c4cf936a71fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_bilbo = tf.keras.models.load_model('best_model_bilbo1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "5239d21e-704f-4a44-bd26-e8fc8bd41726",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 49s 2ms/step\n",
      "Matriz de Confusão\n",
      "[[286579  13242]\n",
      " [ 18415 387427]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9396    0.9558    0.9477    299821\n",
      "           1     0.9670    0.9546    0.9607    405842\n",
      "\n",
      "    accuracy                         0.9551    705663\n",
      "   macro avg     0.9533    0.9552    0.9542    705663\n",
      "weighted avg     0.9553    0.9551    0.9552    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_bilbo.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d0e4be4b-d79e-4131-8ea8-4a21fa71ec47",
   "metadata": {},
   "source": [
    "## Adversarial CNN+LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "3e8cd757-0e06-4f85-97b8-425f984683d0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_1 (Embedding)     (None, 70, 60)            15360     \n",
      "                                                                 \n",
      " conv1d_7 (Conv1D)           (None, 66, 64)            19264     \n",
      "                                                                 \n",
      " max_pooling1d_4 (MaxPoolin  (None, 16, 64)            0         \n",
      " g1D)                                                            \n",
      "                                                                 \n",
      " lstm_1 (LSTM)               (None, 70)                37800     \n",
      "                                                                 \n",
      " dense_3 (Dense)             (None, 1)                 71        \n",
      "                                                                 \n",
      " activation_1 (Activation)   (None, 1)                 0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 72495 (283.18 KB)\n",
      "Trainable params: 72495 (283.18 KB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "hidden_dims=128\n",
    "nb_filter = 64\n",
    "filter_length =5 \n",
    "embedding_vecor_length = 128\n",
    "pool_length=4\n",
    "lstm_output_size = 70\n",
    "\n",
    "\n",
    "\n",
    "model_cnn_lstm = tf.keras.Sequential()\n",
    "model_cnn_lstm.add(layers.Embedding(256, 60, input_length=70))\n",
    "model_cnn_lstm.add(layers.Conv1D(filters=nb_filter, kernel_size=filter_length, activation='relu'))\n",
    "model_cnn_lstm.add(layers.MaxPooling1D(pool_length))\n",
    "model_cnn_lstm.add(layers.LSTM(lstm_output_size))\n",
    "model_cnn_lstm.add(layers.Dense(1))\n",
    "model_cnn_lstm.add(layers.Activation('sigmoid'))\n",
    "\n",
    "model_cnn_lstm.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_cnn_lstm.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "9cb0ed7b-a4fb-4811-b134-0a99ec289e6e",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_cnn_lstm = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model_cnn_lstm1',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "92977199-fb06-4057-a7e6-7b0f7344832c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1490 - accuracy: 0.9422INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 31s 4ms/step - loss: 0.1490 - accuracy: 0.9422 - val_loss: 0.1192 - val_accuracy: 0.9545\n",
      "Epoch 2/10\n",
      "7396/7410 [============================>.] - ETA: 0s - loss: 0.1087 - accuracy: 0.9587INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.1087 - accuracy: 0.9587 - val_loss: 0.1027 - val_accuracy: 0.9610\n",
      "Epoch 3/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0984 - accuracy: 0.9628INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 28s 4ms/step - loss: 0.0984 - accuracy: 0.9628 - val_loss: 0.0960 - val_accuracy: 0.9638\n",
      "Epoch 4/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.0930 - accuracy: 0.9648INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.0930 - accuracy: 0.9648 - val_loss: 0.0922 - val_accuracy: 0.9653\n",
      "Epoch 5/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.0895 - accuracy: 0.9662INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.0895 - accuracy: 0.9662 - val_loss: 0.0917 - val_accuracy: 0.9654\n",
      "Epoch 6/10\n",
      "7399/7410 [============================>.] - ETA: 0s - loss: 0.0865 - accuracy: 0.9674INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.0865 - accuracy: 0.9674 - val_loss: 0.0881 - val_accuracy: 0.9667\n",
      "Epoch 7/10\n",
      "7406/7410 [============================>.] - ETA: 0s - loss: 0.0841 - accuracy: 0.9684INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 28s 4ms/step - loss: 0.0841 - accuracy: 0.9684 - val_loss: 0.0869 - val_accuracy: 0.9674\n",
      "Epoch 8/10\n",
      "7402/7410 [============================>.] - ETA: 0s - loss: 0.0822 - accuracy: 0.9691INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.0822 - accuracy: 0.9690 - val_loss: 0.0859 - val_accuracy: 0.9679\n",
      "Epoch 9/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.0806 - accuracy: 0.9698INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 28s 4ms/step - loss: 0.0806 - accuracy: 0.9698 - val_loss: 0.0837 - val_accuracy: 0.9686\n",
      "Epoch 10/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.0791 - accuracy: 0.9705INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn_lstm1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 29s 4ms/step - loss: 0.0791 - accuracy: 0.9705 - val_loss: 0.0833 - val_accuracy: 0.9689\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x7f0a5440a5f0>"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_cnn_lstm.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best_cnn_lstm],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "e7512e02-17ab-42ad-82b0-16e03683705e",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_cnn_lstm = tf.keras.models.load_model('best_model_cnn_lstm1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "4797b502-507d-4642-b0fb-461b87e7bdad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 20s 882us/step\n",
      "Matriz de Confusão\n",
      "[[291006   8815]\n",
      " [ 13046 392796]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9571    0.9706    0.9638    299821\n",
      "           1     0.9781    0.9679    0.9729    405842\n",
      "\n",
      "    accuracy                         0.9690    705663\n",
      "   macro avg     0.9676    0.9692    0.9684    705663\n",
      "weighted avg     0.9691    0.9690    0.9690    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_cnn_lstm.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c345a68d-0569-4437-952b-b698b3f0a86f",
   "metadata": {},
   "source": [
    "## Adversarial RNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "aa245e9d-e97c-4b41-b631-defcf068aec1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 70, 60)            15360     \n",
      "                                                                 \n",
      " simple_rnn (SimpleRNN)      (None, 128)               24192     \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 128)               0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 1)                 129       \n",
      "                                                                 \n",
      " activation (Activation)     (None, 1)                 0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 39681 (155.00 KB)\n",
      "Trainable params: 39681 (155.00 KB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model_rnn = tf.keras.Sequential()\n",
    "model_rnn.add(layers.Embedding(256, 60, input_length=70))\n",
    "model_rnn.add(layers.SimpleRNN(128))\n",
    "model_rnn.add(layers.Dropout(0.1))\n",
    "model_rnn.add(layers.Dense(1))\n",
    "model_rnn.add(layers.Activation('sigmoid'))\n",
    "\n",
    "\n",
    "model_rnn.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_rnn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "d6e96eae-2421-4c34-9202-5a19b154dc31",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_rnn = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model_rnn1',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "6a556cb8-90ea-4f0e-9700-a11b0c53e56c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 13:58:31.414846: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "2024-04-10 13:58:31.425397: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x63992e5037c0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-10 13:58:31.425413: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-10 13:58:31.429036: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-10 13:58:31.439150: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-10 13:58:31.508802: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.2245 - accuracy: 0.9096INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 139s 19ms/step - loss: 0.2245 - accuracy: 0.9096 - val_loss: 0.1696 - val_accuracy: 0.9346\n",
      "Epoch 2/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1588 - accuracy: 0.9391INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 134s 18ms/step - loss: 0.1588 - accuracy: 0.9391 - val_loss: 0.1479 - val_accuracy: 0.9439\n",
      "Epoch 3/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1418 - accuracy: 0.9459INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 134s 18ms/step - loss: 0.1418 - accuracy: 0.9459 - val_loss: 0.1297 - val_accuracy: 0.9508\n",
      "Epoch 4/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1333 - accuracy: 0.9494INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 132s 18ms/step - loss: 0.1333 - accuracy: 0.9494 - val_loss: 0.1282 - val_accuracy: 0.9516\n",
      "Epoch 5/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1276 - accuracy: 0.9516INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 132s 18ms/step - loss: 0.1276 - accuracy: 0.9516 - val_loss: 0.1228 - val_accuracy: 0.9529\n",
      "Epoch 6/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1231 - accuracy: 0.9535INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 137s 18ms/step - loss: 0.1231 - accuracy: 0.9535 - val_loss: 0.1221 - val_accuracy: 0.9533\n",
      "Epoch 7/10\n",
      "7410/7410 [==============================] - 135s 18ms/step - loss: 0.1195 - accuracy: 0.9549 - val_loss: 0.1367 - val_accuracy: 0.9478\n",
      "Epoch 8/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1164 - accuracy: 0.9562INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 134s 18ms/step - loss: 0.1164 - accuracy: 0.9562 - val_loss: 0.1171 - val_accuracy: 0.9558\n",
      "Epoch 9/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1140 - accuracy: 0.9571INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 132s 18ms/step - loss: 0.1140 - accuracy: 0.9571 - val_loss: 0.1123 - val_accuracy: 0.9580\n",
      "Epoch 10/10\n",
      "7410/7410 [==============================] - ETA: 0s - loss: 0.1113 - accuracy: 0.9582INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_rnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 135s 18ms/step - loss: 0.1113 - accuracy: 0.9582 - val_loss: 0.1132 - val_accuracy: 0.9581\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x701b5eb8bd30>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_rnn.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best_rnn],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "3d9f6997-26c9-4652-a0c4-28f1cbca46eb",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_rnn = tf.keras.models.load_model('best_model_rnn1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5a136c71-596a-43f3-a0c0-2deea09953e7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 87s 4ms/step\n",
      "Matriz de Confusão\n",
      "[[289658  10163]\n",
      " [ 19697 386145]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9363    0.9661    0.9510    299821\n",
      "           1     0.9744    0.9515    0.9628    405842\n",
      "\n",
      "    accuracy                         0.9577    705663\n",
      "   macro avg     0.9553    0.9588    0.9569    705663\n",
      "weighted avg     0.9582    0.9577    0.9578    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_rnn.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82f9022e-8d2f-4913-bc54-38a6c922285e",
   "metadata": {},
   "source": [
    "## Adversarial CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "408ab659-6ffb-46b8-a442-1c8c395f120f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding (Embedding)       (None, 70, 60)            15360     \n",
      "                                                                 \n",
      " conv1d (Conv1D)             (None, 66, 64)            19264     \n",
      "                                                                 \n",
      " global_max_pooling1d (Glob  (None, 64)                0         \n",
      " alMaxPooling1D)                                                 \n",
      "                                                                 \n",
      " dense (Dense)               (None, 128)               8320      \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 128)               0         \n",
      "                                                                 \n",
      " activation (Activation)     (None, 128)               0         \n",
      "                                                                 \n",
      " dense_1 (Dense)             (None, 1)                 129       \n",
      "                                                                 \n",
      " activation_1 (Activation)   (None, 1)                 0         \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 43073 (168.25 KB)\n",
      "Trainable params: 43073 (168.25 KB)\n",
      "Non-trainable params: 0 (0.00 Byte)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "hidden_dims=128\n",
    "nb_filter = 64\n",
    "filter_length =5 \n",
    "embedding_vecor_length = 128\n",
    "pool_length=4\n",
    "lstm_output_size = 70\n",
    "\n",
    "model_cnn = tf.keras.Sequential()\n",
    "model_cnn.add(layers.Embedding(256, 60, input_length=70))\n",
    "model_cnn.add(layers.Conv1D(filters=nb_filter, kernel_size=filter_length, activation='relu'))\n",
    "model_cnn.add(layers.GlobalMaxPooling1D())\n",
    "model_cnn.add(layers.Dense(hidden_dims))\n",
    "model_cnn.add(layers.Dropout(0.2))\n",
    "model_cnn.add(layers.Activation('relu'))\n",
    "model_cnn.add(layers.Dense(1))\n",
    "model_cnn.add(layers.Activation('sigmoid'))\n",
    "\n",
    "model_cnn.compile(optimizer = opt, loss = 'binary_crossentropy', metrics='accuracy')\n",
    "model_cnn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "eba98469-3854-4804-a9cd-31bee3f8b9af",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_cnn = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model_cnn1',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "fd4df754-4526-400e-8360-30f7208456e5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-10 14:31:14.405319: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-10 14:31:14.541658: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "2024-04-10 14:31:14.544370: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x5b95064e6950 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-10 14:31:14.544384: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-10 14:31:14.548241: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-10 14:31:14.629067: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - ETA: 0s - loss: 0.2187 - accuracy: 0.9097INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 23s 3ms/step - loss: 0.2187 - accuracy: 0.9097 - val_loss: 0.1541 - val_accuracy: 0.9379\n",
      "Epoch 2/10\n",
      "7395/7410 [============================>.] - ETA: 0s - loss: 0.1407 - accuracy: 0.9437INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1407 - accuracy: 0.9437 - val_loss: 0.1301 - val_accuracy: 0.9479\n",
      "Epoch 3/10\n",
      "7406/7410 [============================>.] - ETA: 0s - loss: 0.1238 - accuracy: 0.9506INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1238 - accuracy: 0.9506 - val_loss: 0.1186 - val_accuracy: 0.9529\n",
      "Epoch 4/10\n",
      "7407/7410 [============================>.] - ETA: 0s - loss: 0.1152 - accuracy: 0.9542INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1152 - accuracy: 0.9542 - val_loss: 0.1127 - val_accuracy: 0.9553\n",
      "Epoch 5/10\n",
      "7408/7410 [============================>.] - ETA: 0s - loss: 0.1095 - accuracy: 0.9567INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1095 - accuracy: 0.9567 - val_loss: 0.1097 - val_accuracy: 0.9564\n",
      "Epoch 6/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1052 - accuracy: 0.9586INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1052 - accuracy: 0.9586 - val_loss: 0.1057 - val_accuracy: 0.9584\n",
      "Epoch 7/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1022 - accuracy: 0.9599INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.1022 - accuracy: 0.9599 - val_loss: 0.1035 - val_accuracy: 0.9591\n",
      "Epoch 8/10\n",
      "7403/7410 [============================>.] - ETA: 0s - loss: 0.0997 - accuracy: 0.9609INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.0997 - accuracy: 0.9609 - val_loss: 0.1021 - val_accuracy: 0.9597\n",
      "Epoch 9/10\n",
      "7388/7410 [============================>.] - ETA: 0s - loss: 0.0974 - accuracy: 0.9618INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.0974 - accuracy: 0.9618 - val_loss: 0.1007 - val_accuracy: 0.9601\n",
      "Epoch 10/10\n",
      "7397/7410 [============================>.] - ETA: 0s - loss: 0.0956 - accuracy: 0.9627INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_cnn1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 18s 2ms/step - loss: 0.0956 - accuracy: 0.9627 - val_loss: 0.1003 - val_accuracy: 0.9605\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x7dd8674c2e00>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_cnn.fit(X_train, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best_cnn],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "1e5716c0-b7f3-4407-9300-98a6f936d252",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_cnn = tf.keras.models.load_model('best_model_cnn1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "bf2e126c-8ad9-404b-8e51-451cf155c40c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 13s 585us/step\n",
      "Matriz de Confusão\n",
      "[[290088   9733]\n",
      " [ 17524 388318]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9430    0.9675    0.9551    299821\n",
      "           1     0.9755    0.9568    0.9661    405842\n",
      "\n",
      "    accuracy                         0.9614    705663\n",
      "   macro avg     0.9593    0.9622    0.9606    705663\n",
      "weighted avg     0.9617    0.9614    0.9614    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_cnn.predict(X_test)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "799bf26d-c07a-409e-b3dd-946d24d9c079",
   "metadata": {},
   "source": [
    "## DistilBERT Tokenizer + DNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7881e841-6a84-4c29-9933-a824648bbbbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_test = df_test.reset_index(drop=True)\n",
    "df_train = df_train.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "15ca9f57-c7dd-4339-8b32-e2b5c9cd3610",
   "metadata": {},
   "outputs": [],
   "source": [
    "import transformers\n",
    "tokenizer = transformers.DistilBertTokenizer.from_pretrained('distilbert-base-uncased')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "a4698e89-df6a-42dc-9f3c-2d3937e0d845",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = df_train['domain']\n",
    "y_train = df_train['label']\n",
    "y_train = y_train.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "b16ee863-f537-4143-8839-460cf3dd9cd8",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test = df_test['domain']\n",
    "y_test = df_test['label']\n",
    "y_test = y_test.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "6abe2825-5759-4699-a025-2c8111d2ccbb",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_tk = [tokenizer(text, padding='max_length', max_length = 10, truncation=True)['input_ids'] for text in X_train]\n",
    "X_train_tk = np.array(X_train_tk, dtype='int32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "1b2ebccf-7130-44b1-b8ed-de4b090277e5",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test_tk = [tokenizer(text, padding='max_length', max_length = 10, truncation=True)['input_ids'] for text in X_test]\n",
    "X_test_tk = np.array(X_test_tk, dtype='int32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4ed623d8-9184-42bf-8dec-7c83cadac373",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-13 16:26:56.179131: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:56.201720: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:56.201941: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:56.202871: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:56.203005: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:56.203077: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:57.206716: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:57.206879: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:57.206961: I tensorflow/compiler/xla/stream_executor/cuda/cuda_gpu_executor.cc:995] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero. See more at https://github.com/torvalds/linux/blob/v6.0/Documentation/ABI/testing/sysfs-bus-pci#L344-L355\n",
      "2024-04-13 16:26:57.207037: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1639] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 1745 MB memory:  -> device: 0, name: NVIDIA GeForce RTX 3050 Laptop GPU, pci bus id: 0000:01:00.0, compute capability: 8.6\n",
      "2024-04-13 16:26:58.220136: I tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:606] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.\n",
      "Some weights of the PyTorch model were not used when initializing the TF 2.0 model TFDistilBertModel: ['vocab_projector.bias', 'vocab_transform.weight', 'vocab_transform.bias', 'vocab_layer_norm.bias', 'vocab_layer_norm.weight']\n",
      "- This IS expected if you are initializing TFDistilBertModel from a PyTorch model trained on another task or with another architecture (e.g. initializing a TFBertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing TFDistilBertModel from a PyTorch model that you expect to be exactly identical (e.g. initializing a TFBertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "All the weights of TFDistilBertModel were initialized from the PyTorch model.\n",
      "If your task is similar to the task the model of the checkpoint was trained on, you can already use TFDistilBertModel for predictions without further training.\n"
     ]
    }
   ],
   "source": [
    "config = transformers.DistilBertConfig(dropout=0.2, attention_dropout=0.2)\n",
    "dbert_tf = transformers.TFDistilBertModel.from_pretrained('distilbert-base-uncased', config=config, trainable=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "bc0021db-fdfd-4067-b401-bdc9ce7a74db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  101,  1023,  2278,  2546,  2575,  2497,  2683, 24594,  2683,\n",
       "         102], dtype=int32)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_tk[2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "1d7c146d-ab7d-4804-9d1d-253e50bc3b33",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"ClassificationModelTF\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " input_token (InputLayer)    [(None, 10)]              0         \n",
      "                                                                 \n",
      " tf_distil_bert_model (TFDi  TFBaseModelOutput(last_   66362880  \n",
      " stilBertModel)              hidden_state=(None, 10,             \n",
      "                              768),                              \n",
      "                              hidden_states=None, at             \n",
      "                             tentions=None)                      \n",
      "                                                                 \n",
      " tf.__operators__.getitem (  (None, 768)               0         \n",
      " SlicingOpLambda)                                                \n",
      "                                                                 \n",
      " dropout (Dropout)           (None, 768)               0         \n",
      "                                                                 \n",
      " dense (Dense)               (None, 64)                49216     \n",
      "                                                                 \n",
      " classification (Dense)      (None, 1)                 65        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 66412161 (253.34 MB)\n",
      "Trainable params: 49281 (192.50 KB)\n",
      "Non-trainable params: 66362880 (253.15 MB)\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "input_ids_in = layers.Input(shape=(10,), name='input_token', dtype='int32')\n",
    "\n",
    "x = dbert_tf(input_ids=input_ids_in)[0][:,0,:]\n",
    "x = layers.Dropout(0.2, name='dropout')(x)\n",
    "x = layers.Dense(64, activation='relu', name='dense')(x)\n",
    "x = layers.Dense(1, activation='sigmoid', name='classification')(x)\n",
    "\n",
    "model_bert = tf.keras.Model(inputs=input_ids_in, outputs = x, name='ClassificationModelTF')\n",
    "\n",
    "model_bert.compile(optimizer='adam',loss='binary_crossentropy', metrics='accuracy')\n",
    "model_bert.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "b40fe237-d365-42d1-9e95-5456ca4d219c",
   "metadata": {},
   "outputs": [],
   "source": [
    "best_bert = tf.keras.callbacks.ModelCheckpoint(\n",
    "    filepath='best_model_bert1',\n",
    "    save_weights_only=False,\n",
    "    monitor='val_accuracy',\n",
    "    mode='max',\n",
    "    save_best_only=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "9d880f67-05b0-4a20-91f0-5fe16f2d0c40",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-04-13 16:27:03.917428: I tensorflow/compiler/xla/service/service.cc:168] XLA service 0x70841c0509f0 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
      "2024-04-13 16:27:03.917446: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): NVIDIA GeForce RTX 3050 Laptop GPU, Compute Capability 8.6\n",
      "2024-04-13 16:27:03.921262: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:255] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
      "2024-04-13 16:27:03.932295: I tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:432] Loaded cuDNN version 8902\n",
      "2024-04-13 16:27:04.017982: I ./tensorflow/compiler/jit/device_compiler.h:186] Compiled cluster using XLA!  This line is logged at most once for the lifetime of the process.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2349 - accuracy: 0.8985WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n",
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n",
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n",
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n",
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n",
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n",
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 525s 70ms/step - loss: 0.2349 - accuracy: 0.8985 - val_loss: 0.1666 - val_accuracy: 0.9340\n",
      "Epoch 2/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2090 - accuracy: 0.9104WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 524s 71ms/step - loss: 0.2090 - accuracy: 0.9104 - val_loss: 0.1523 - val_accuracy: 0.9383\n",
      "Epoch 3/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.2024 - accuracy: 0.9132WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 524s 71ms/step - loss: 0.2024 - accuracy: 0.9132 - val_loss: 0.1506 - val_accuracy: 0.9403\n",
      "Epoch 4/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1974 - accuracy: 0.9156WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 524s 71ms/step - loss: 0.1974 - accuracy: 0.9156 - val_loss: 0.1464 - val_accuracy: 0.9410\n",
      "Epoch 5/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1945 - accuracy: 0.9166WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 524s 71ms/step - loss: 0.1945 - accuracy: 0.9166 - val_loss: 0.1415 - val_accuracy: 0.9431\n",
      "Epoch 6/10\n",
      "7410/7410 [==============================] - 515s 69ms/step - loss: 0.1920 - accuracy: 0.9178 - val_loss: 0.1502 - val_accuracy: 0.9378\n",
      "Epoch 7/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1903 - accuracy: 0.9182WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 525s 71ms/step - loss: 0.1903 - accuracy: 0.9182 - val_loss: 0.1383 - val_accuracy: 0.9447\n",
      "Epoch 8/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1887 - accuracy: 0.9195WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 525s 71ms/step - loss: 0.1887 - accuracy: 0.9195 - val_loss: 0.1370 - val_accuracy: 0.9453\n",
      "Epoch 9/10\n",
      "7410/7410 [==============================] - 514s 69ms/step - loss: 0.1879 - accuracy: 0.9195 - val_loss: 0.1394 - val_accuracy: 0.9435\n",
      "Epoch 10/10\n",
      "7409/7410 [============================>.] - ETA: 0s - loss: 0.1865 - accuracy: 0.9201WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c10bb0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c13730>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694c263b0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbd030>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cbfc70>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Skipping full serialization of Keras layer <keras.src.layers.regularization.dropout.Dropout object at 0x708694cca8f0>, because it is not built.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Assets written to: best_model_bert1/assets\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7410/7410 [==============================] - 524s 71ms/step - loss: 0.1865 - accuracy: 0.9201 - val_loss: 0.1343 - val_accuracy: 0.9464\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x70869df89ba0>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model_bert.fit(X_train_tk, y_train,\n",
    "              epochs=10,\n",
    "              validation_split = 0.1,\n",
    "              callbacks = [best_bert],\n",
    "              batch_size=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "cf9210fa-9665-49c8-9972-99d2f42eee60",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22052/22052 [==============================] - 374s 17ms/step\n",
      "Matriz de Confusão\n",
      "[[282834  16987]\n",
      " [ 20793 385049]]\n",
      "Relatório de Classificação\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0     0.9315    0.9433    0.9374    299821\n",
      "           1     0.9577    0.9488    0.9532    405842\n",
      "\n",
      "    accuracy                         0.9465    705663\n",
      "   macro avg     0.9446    0.9461    0.9453    705663\n",
      "weighted avg     0.9466    0.9465    0.9465    705663\n",
      "\n"
     ]
    }
   ],
   "source": [
    "predictions = model_bert.predict(X_test_tk)\n",
    "predictions = np.where(predictions > 0.5, 1, 0)\n",
    "\n",
    "print('Matriz de Confusão')\n",
    "print(confusion_matrix(y_test,predictions))\n",
    "\n",
    "print('Relatório de Classificação')\n",
    "print(classification_report(y_test,predictions, digits=4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "530eff60-5b04-456f-9bda-2ef825c146ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_bert.save_weights('best_model_bert1_weights')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}