07. Classify Article.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Classify Articles:Three Steps\n",
    "- Start with a corpus of articles\n",
    "- Identify underlying themes using kmeans++\n",
    "- Assign themes to new article\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#We have 2804 articles under a folder, read and store them in a List of articles\n",
    "#In case error reading so many files in Jupyter\n",
    "#Step1:On CMD Prompt: $ jupyter notebook --generate-config\n",
    "#Step2: Open the file displayed above and search for c.NotebookApp.iopub_data_rate_limit\n",
    "#Step3: Comment out the line c.NotebookApp.iopub_data_rate_limit = 1000000 and increase one more zero\n",
    "\n",
    "articles = []\n",
    "for i in range(0, 2804):\n",
    "    with open('Data/NLP/%s.txt'%i, 'r') as file:\n",
    "        text = file.read()\n",
    "        articles.append(text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#print(articles)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<2804x13220 sparse matrix of type '<class 'numpy.float64'>'\n",
       "\twith 280835 stored elements in Compressed Sparse Row format>"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Convert text to TF-IDF representation\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "\n",
    "#By setting stopwords to relevant language, it will ignore the stopwords\n",
    "vectorizer = TfidfVectorizer(max_df=0.5,min_df=2,stop_words='english')\n",
    "\n",
    "X = vectorizer.fit_transform(articles)\n",
    "X\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "13220"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(vectorizer.vocabulary_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster = 2 wcss = 2678.878747558282\n",
      "Cluster = 2 silhouette = 0.023648805049372647\n",
      "Cluster = 3 wcss = 2655.0533240237596\n",
      "Cluster = 3 silhouette = 0.014171285899624882\n",
      "Cluster = 4 wcss = 2637.8093069612523\n",
      "Cluster = 4 silhouette = 0.015319906468816324\n",
      "Cluster = 5 wcss = 2617.768022116416\n",
      "Cluster = 5 silhouette = 0.017847629183682347\n",
      "Cluster = 6 wcss = 2601.2448421961967\n",
      "Cluster = 6 silhouette = 0.019611983728491836\n",
      "Cluster = 7 wcss = 2585.2348984784826\n",
      "Cluster = 7 silhouette = 0.02123455725552626\n",
      "Cluster = 8 wcss = 2577.7206582866697\n",
      "Cluster = 8 silhouette = 0.021239426711191337\n",
      "Cluster = 9 wcss = 2562.406257499248\n",
      "Cluster = 9 silhouette = 0.022787451614507136\n",
      "Cluster = 10 wcss = 2556.24989937093\n",
      "Cluster = 10 silhouette = 0.023200325194040524\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Using the elbow method to find the optimal number of clusters\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.metrics import silhouette_score\n",
    "\n",
    "wcss = []\n",
    "silhouette = []\n",
    "for i in range(2, 11):\n",
    "    k_means = KMeans(n_clusters = i, init = 'k-means++', random_state = 42)\n",
    "    k_means.fit(X)\n",
    "    print(\"Cluster = {0} wcss = {1}\".format(i, k_means.inertia_))\n",
    "    wcss.append(k_means.inertia_)\n",
    "    pred = k_means.predict(X)\n",
    "    print(\"Cluster = {0} silhouette = {1}\".format(i, silhouette_score(X, pred)))\n",
    "    silhouette.append(silhouette_score(X, pred))\n",
    "plt.plot(range(2, 11), wcss)\n",
    "plt.title('Elbow - Method')\n",
    "plt.xlabel('Number of Clusters')\n",
    "plt.ylabel('WCSS')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(range(2, 11), silhouette)\n",
    "plt.title('Silhouette')\n",
    "plt.xlabel('Number of Clusters')\n",
    "plt.ylabel('Silhouette')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initialization complete\n",
      "Iteration  0, inertia 5307.140\n",
      "Iteration  1, inertia 2687.485\n",
      "Iteration  2, inertia 2677.779\n",
      "Iteration  3, inertia 2672.270\n",
      "Iteration  4, inertia 2669.639\n",
      "Iteration  5, inertia 2669.005\n",
      "Iteration  6, inertia 2668.738\n",
      "Iteration  7, inertia 2668.529\n",
      "Iteration  8, inertia 2668.343\n",
      "Iteration  9, inertia 2668.186\n",
      "Iteration 10, inertia 2668.023\n",
      "Iteration 11, inertia 2667.812\n",
      "Iteration 12, inertia 2667.674\n",
      "Iteration 13, inertia 2667.510\n",
      "Iteration 14, inertia 2667.433\n",
      "Iteration 15, inertia 2667.402\n",
      "Iteration 16, inertia 2667.387\n",
      "Iteration 17, inertia 2667.381\n",
      "Iteration 18, inertia 2667.376\n",
      "Iteration 19, inertia 2667.373\n",
      "Converged at iteration 19: center shift 0.000000e+00 within tolerance 7.307995e-09\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "KMeans(algorithm='auto', copy_x=True, init='k-means++', max_iter=100,\n",
       "       n_clusters=3, n_init=1, n_jobs=None, precompute_distances='auto',\n",
       "       random_state=None, tol=0.0001, verbose=True)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#X[0]\n",
    "num_clus = 3\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "#max_iter means if algorithm don't converge even after this no# of iteration, it will stop\n",
    "km = KMeans(n_clusters = num_clus, init = 'k-means++', max_iter = 100, n_init = 1, verbose = True)\n",
    "km.fit(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([0, 1, 2]), array([ 452, 1943,  409], dtype=int64))"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "#km.lables_ is an attribute of K-means object. unique function will print the unique clusters name with each article have been assign one cluster\n",
    "np.unique(km.labels_, return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Collect articles in each cluster\n",
    "cluster={}\n",
    "for i,clus_no in enumerate(km.labels_):\n",
    "        indiv_doc = articles[i]\n",
    "        if clus_no not in cluster.keys():\n",
    "            cluster[clus_no] = indiv_doc\n",
    "        else:\n",
    "            cluster[clus_no] += indiv_doc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.tokenize import sent_tokenize, word_tokenize\n",
    "from nltk.corpus import stopwords\n",
    "from nltk.probability import FreqDist\n",
    "from collections import defaultdict\n",
    "from string import punctuation\n",
    "from heapq import nlargest\n",
    "import nltk\n",
    "\n",
    "_stopwords = set(stopwords.words('english') + list(punctuation)+[\"month\",\"year\",\"day\",\"'s\",\"''\",\"``\"])\n",
    "\n",
    "#Top keywords in each cluster and their count\n",
    "keywords = {}\n",
    "counts={}\n",
    "for clus_no in range(num_clus):\n",
    "    word_sent = word_tokenize(cluster[clus_no].lower())\n",
    "    word_sent=[word for word in word_sent if word not in _stopwords]\n",
    "    freq = FreqDist(word_sent)\n",
    "    keywords[clus_no] = nlargest(100, freq, key=freq.get)\n",
    "    counts[clus_no]=freq\n",
    "\n",
    "\n",
    "                                                                 \n",
    "                                        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Find unique keywords in each cluster\n",
    "unique_keys={}\n",
    "for clus_no in range(num_clus):\n",
    "    rest_clusters=list(set(range(num_clus))-set([clus_no]))\n",
    "    #collect all the keywords from other clusters\n",
    "    for i in range(num_clus - 1):\n",
    "        if i == 0:\n",
    "            keywords_other_clusters=set(keywords[rest_clusters[i]])\n",
    "        else:\n",
    "            keywords_other_clusters=keywords_other_clusters.union(set(keywords[rest_clusters[i]]))\n",
    "    #Take out these keywords from the whole set to get unique keywords\n",
    "    unique=set(keywords[clus_no])-keywords_other_clusters\n",
    "    #Store top 10 keywords \n",
    "    unique_keys[clus_no]=nlargest(10, unique, key=counts[clus_no].get)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: ['quarter',\n",
       "  'share',\n",
       "  'profit',\n",
       "  'rose',\n",
       "  'analysts',\n",
       "  'earnings',\n",
       "  'fell',\n",
       "  'cents',\n",
       "  'net',\n",
       "  'per'],\n",
       " 1: ['ads',\n",
       "  'video',\n",
       "  'products',\n",
       "  'internet',\n",
       "  'use',\n",
       "  'product',\n",
       "  'search',\n",
       "  'apps',\n",
       "  'pay',\n",
       "  'buy'],\n",
       " 2: ['uber',\n",
       "  'round',\n",
       "  'valuation',\n",
       "  'funding',\n",
       "  'raised',\n",
       "  'private',\n",
       "  'capital',\n",
       "  'lyft',\n",
       "  'investment',\n",
       "  'drivers']}"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "unique_keys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Article Category = online_deal\n"
     ]
    }
   ],
   "source": [
    "name_article = {0: \"share_market\", 1: \"online_deal\", 2: \"startup_investment\"}\n",
    "\n",
    "article = \"US-based Facebook has agreed to buy 9.99% equity stake in oil-to-retail conglomerate Reliance Industries Ltd’s telecommunications arm Reliance Jio for Rs 43,574 crore. While the Mukesh Ambani-led company said that Facebook and Jio will together further India’s digital economy, the deal in itself has retail at its core, as both Facebook and Reliance Jio eye a huge market of 1.3 billion Indians and Indian businesses. “Our focus will be India’s 60 million micro, small and medium businesses, 120 million farmers, 30 million small merchants and millions of small and medium enterprises in the informal sector,” Reliance Jio said in the statement announcing the deal. In fact, RIL has been building up to become a retail giant with several acquisitions and deals in the past as well.\"\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "#here K(n_neighbors)=5 by default\n",
    "classifier = KNeighborsClassifier(n_neighbors=num_clus)\n",
    "\n",
    "#training phase\n",
    "classifier.fit(X,km.labels_)\n",
    "test=vectorizer.transform([article.encode('ascii',errors='ignore')])\n",
    "cluster_no = classifier.predict(test)\n",
    "print(\"Article Category = %s\" % name_article[list(cluster_no)[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}