example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Example for toy data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create toy data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(12345)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(x):\n",
    "    return 1.0 / (1.0 + np.exp(-x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.DataFrame(np.random.rand(500, 5), columns=list('abcde'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "W = np.array([[1, -2, 3, -2, 1], [-2, 1, 3, -2, 1]])\n",
    "b = np.array([0.2, -0.1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "margins = df.dot(W.transpose()).add(b)\n",
    "p = margins[0].map(sigmoid)\n",
    "l = np.exp(margins[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['time'] = l.map(lambda l:np.clip(np.random.exponential(l, 1)[0], 0, 1))\n",
    "df['label'] = pd.concat([p, l], axis=1).apply(lambda r:np.round(r[0] * (1. - np.exp(-r[1]))), axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_parquet('toy.parquet')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train & Test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyspark.ml import Pipeline\n",
    "from pyspark.ml.feature import RFormula\n",
    "from pyspark.ml.evaluation import BinaryClassificationEvaluator\n",
    "from dfm.classification import DelayedFeedbackClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "raw = sqlContext.read.parquet('toy.parquet')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "root\n",
      " |-- a: double (nullable = true)\n",
      " |-- b: double (nullable = true)\n",
      " |-- c: double (nullable = true)\n",
      " |-- d: double (nullable = true)\n",
      " |-- e: double (nullable = true)\n",
      " |-- time: double (nullable = true)\n",
      " |-- label: double (nullable = true)\n",
      " |-- __index_level_0__: long (nullable = true)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "raw.printSchema()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "train, test = raw.randomSplit([0.9, 0.1], seed=12345)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "formula = RFormula(formula='label ~ a + b + c + d + e')\n",
    "dfc = DelayedFeedbackClassifier(timeCol='time', regParam=0.01)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "pipeline = Pipeline(stages=[formula, dfc])\n",
    "model = pipeline.fit(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9779310344827586"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "eval = BinaryClassificationEvaluator()\n",
    "eval.evaluate(predictions)"
   ]
  }
 ],
 "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}