TFIDF_classification_reviews.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## TFIDF classification\n",
    "\n",
    "#### Dataset: [Kaggle Bag of Words meets Bag of Popcorns](https://www.kaggle.com/c/word2vec-nlp-tutorial/data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from nltk.corpus import stopwords\n",
    "import re\n",
    "from bs4 import BeautifulSoup\n",
    "import numpy as np\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from scipy import sparse\n",
    "import matplotlib.pyplot as plt\n",
    "% matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train = pd.read_csv(\"labeledTrainData.tsv\", header=0, delimiter=\"\\t\", quoting=3, encoding='utf-8')\n",
    "test = pd.read_csv(\"testData.tsv\", header=0, delimiter=\"\\t\", quoting=3, encoding='utf-8')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Text cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "stop = set(stopwords.words(\"english\"))\n",
    "def clean_text(text):\n",
    "    review_text = BeautifulSoup(text, \"html5lib\").get_text()\n",
    "    letter_only = re.sub(\"[^a-zA-Z]\",\" \", review_text)\n",
    "    words = letter_only.split()\n",
    "    clean_text = [w for w in words if not w in stop]\n",
    "    return(\" \".join(clean_text))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "u'\"I really wanted to like this film, but the story is ridicules. I don\\'t want to spoil this film, - don\\'t worry right from the begin you know something bad is going to happen - but here\\'s an example of how sloppy this film was put together. The Cowboy and \\\\\"Twig\\\\\" ride up the ridge. The Cowboy has a handle bar mustache. The Cowboy and \\\\\"Twig\\\\\" get into a shoot out and race half way down the ridge. The Cowboy is clean shaven through out the rest of the film. Sometime between the gun fight and the ride down the mountain the cowboy has had time to shave, in dark, on the back of a horse.<br /><br />To be fair, the acting by the four main characters is solid.\"'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train['review'][69]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "u'I really wanted like film story ridicules I want spoil film worry right begin know something bad going happen example sloppy film put together The Cowboy Twig ride ridge The Cowboy handle bar mustache The Cowboy Twig get shoot race half way ridge The Cowboy clean shaven rest film Sometime gun fight ride mountain cowboy time shave dark back horse To fair acting four main characters solid'"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clean_text(train['review'][69])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1000\n",
      "2000\n",
      "3000\n",
      "4000\n",
      "5000\n",
      "6000\n",
      "7000\n",
      "8000\n",
      "9000\n",
      "10000\n",
      "11000\n",
      "12000\n",
      "13000\n",
      "14000\n",
      "15000\n",
      "16000\n",
      "17000\n",
      "18000\n",
      "19000\n",
      "20000\n",
      "21000\n",
      "22000\n",
      "23000\n",
      "24000\n",
      "25000\n",
      "1000\n",
      "2000\n",
      "3000\n",
      "4000\n",
      "5000\n",
      "6000\n",
      "7000\n",
      "8000\n",
      "9000\n",
      "10000\n",
      "11000\n",
      "12000\n",
      "13000\n",
      "14000\n",
      "15000\n",
      "16000\n",
      "17000\n",
      "18000\n",
      "19000\n",
      "20000\n",
      "21000\n",
      "22000\n",
      "23000\n",
      "24000\n",
      "25000\n"
     ]
    }
   ],
   "source": [
    "clean_train_reviews = []\n",
    "clean_test_reviews = []\n",
    "\n",
    "for i in xrange(train[\"review\"].size):\n",
    "    if (i+1)%1000 == 0:\n",
    "        print i+1\n",
    "    clean_train_reviews.append(clean_text(train[\"review\"][i]))\n",
    "\n",
    "for i in xrange(test[\"review\"].size):\n",
    "    if (i+1)%1000 == 0:\n",
    "        print i+1\n",
    "    clean_test_reviews.append(clean_text(test[\"review\"][i]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Initializing TFIDF Vectorizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tfidf = TfidfVectorizer(min_df=3, \n",
    "                        max_features=1000, \n",
    "                        strip_accents='unicode', \n",
    "                        analyzer='word', \n",
    "                        ngram_range=(1, 2), \n",
    "                        use_idf=1, \n",
    "                        smooth_idf=1, \n",
    "                        sublinear_tf=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_data_features = tfidf.fit_transform(clean_train_reviews)\n",
    "test_data_features = tfidf.transform(clean_test_reviews)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25000, 1000)\n",
      "(25000, 1000)\n"
     ]
    }
   ],
   "source": [
    "print train_data_features.shape\n",
    "print test_data_features.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_labels = train[\"sentiment\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Activation\n",
    "from keras.utils import np_utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_labels = np_utils.to_categorical(train_labels, 2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baseline Model (Softmax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(input_shape=(1000,), units=2))\n",
    "model.add(Activation('softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 20000 samples, validate on 5000 samples\n",
      "Epoch 1/10\n",
      "0s - loss: 0.6864 - acc: 0.5837 - binary_accuracy: 0.5837 - val_loss: 0.6826 - val_acc: 0.6302 - val_binary_accuracy: 0.6302\n",
      "Epoch 2/10\n",
      "0s - loss: 0.6783 - acc: 0.6667 - binary_accuracy: 0.6667 - val_loss: 0.6749 - val_acc: 0.7022 - val_binary_accuracy: 0.7022\n",
      "Epoch 3/10\n",
      "0s - loss: 0.6706 - acc: 0.7205 - binary_accuracy: 0.7205 - val_loss: 0.6676 - val_acc: 0.7456 - val_binary_accuracy: 0.7456\n",
      "Epoch 4/10\n",
      "0s - loss: 0.6631 - acc: 0.7586 - binary_accuracy: 0.7586 - val_loss: 0.6608 - val_acc: 0.7574 - val_binary_accuracy: 0.7574\n",
      "Epoch 5/10\n",
      "1s - loss: 0.6559 - acc: 0.7776 - binary_accuracy: 0.7776 - val_loss: 0.6539 - val_acc: 0.7824 - val_binary_accuracy: 0.7824\n",
      "Epoch 6/10\n",
      "1s - loss: 0.6490 - acc: 0.7925 - binary_accuracy: 0.7925 - val_loss: 0.6472 - val_acc: 0.7950 - val_binary_accuracy: 0.7950\n",
      "Epoch 7/10\n",
      "1s - loss: 0.6423 - acc: 0.8019 - binary_accuracy: 0.8019 - val_loss: 0.6410 - val_acc: 0.8028 - val_binary_accuracy: 0.8028\n",
      "Epoch 8/10\n",
      "1s - loss: 0.6358 - acc: 0.8092 - binary_accuracy: 0.8092 - val_loss: 0.6348 - val_acc: 0.8084 - val_binary_accuracy: 0.8084\n",
      "Epoch 9/10\n",
      "0s - loss: 0.6295 - acc: 0.8144 - binary_accuracy: 0.8144 - val_loss: 0.6288 - val_acc: 0.8150 - val_binary_accuracy: 0.8150\n",
      "Epoch 10/10\n",
      "0s - loss: 0.6234 - acc: 0.8175 - binary_accuracy: 0.8175 - val_loss: 0.6232 - val_acc: 0.8156 - val_binary_accuracy: 0.8156\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='binary_crossentropy', \n",
    "              optimizer='sgd', \n",
    "              metrics=['accuracy', 'binary_accuracy'])\n",
    "\n",
    "history = model.fit(train_data_features.toarray(), train_labels, \n",
    "                    validation_split=0.2,\n",
    "                    verbose=2)\n",
    "# Even at 1,000 epochs the max validation accuracy it gives is 0.82"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['acc', 'loss', 'val_binary_accuracy', 'binary_accuracy', 'val_acc', 'val_loss']\n"
     ]
    }
   ],
   "source": [
    "print history.history.keys()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train and validation accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f889141b790>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nxhjTKAsLY4wxjfr/E6a9RwXDsLwAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8891df4e90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train and Validation loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f889141bd50>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJdlG0AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f889144f050>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## FC + RELU + RELU + FC + Softmax (MLP)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(input_shape=(1000,), units=512))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Activation('relu'))\n",
    "model.add(Dense(2))\n",
    "model.add(Activation('softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 20000 samples, validate on 5000 samples\n",
      "Epoch 1/50\n",
      "4s - loss: 0.6899 - acc: 0.5603 - binary_accuracy: 0.5603 - val_loss: 0.6854 - val_acc: 0.6160 - val_binary_accuracy: 0.6160\n",
      "Epoch 2/50\n",
      "5s - loss: 0.6805 - acc: 0.6791 - binary_accuracy: 0.6791 - val_loss: 0.6758 - val_acc: 0.7002 - val_binary_accuracy: 0.7002\n",
      "Epoch 3/50\n",
      "5s - loss: 0.6686 - acc: 0.7419 - binary_accuracy: 0.7419 - val_loss: 0.6622 - val_acc: 0.7628 - val_binary_accuracy: 0.7628\n",
      "Epoch 4/50\n",
      "4s - loss: 0.6525 - acc: 0.7783 - binary_accuracy: 0.7783 - val_loss: 0.6443 - val_acc: 0.7892 - val_binary_accuracy: 0.7892\n",
      "Epoch 5/50\n",
      "4s - loss: 0.6311 - acc: 0.7973 - binary_accuracy: 0.7973 - val_loss: 0.6207 - val_acc: 0.8004 - val_binary_accuracy: 0.8004\n",
      "Epoch 6/50\n",
      "4s - loss: 0.6039 - acc: 0.8067 - binary_accuracy: 0.8067 - val_loss: 0.5920 - val_acc: 0.8120 - val_binary_accuracy: 0.8120\n",
      "Epoch 7/50\n",
      "5s - loss: 0.5722 - acc: 0.8158 - binary_accuracy: 0.8158 - val_loss: 0.5616 - val_acc: 0.8102 - val_binary_accuracy: 0.8102\n",
      "Epoch 8/50\n",
      "6s - loss: 0.5380 - acc: 0.8229 - binary_accuracy: 0.8229 - val_loss: 0.5267 - val_acc: 0.8224 - val_binary_accuracy: 0.8224\n",
      "Epoch 9/50\n",
      "5s - loss: 0.5043 - acc: 0.8278 - binary_accuracy: 0.8278 - val_loss: 0.4956 - val_acc: 0.8270 - val_binary_accuracy: 0.8270\n",
      "Epoch 10/50\n",
      "5s - loss: 0.4734 - acc: 0.8337 - binary_accuracy: 0.8337 - val_loss: 0.4676 - val_acc: 0.8304 - val_binary_accuracy: 0.8304\n",
      "Epoch 11/50\n",
      "4s - loss: 0.4464 - acc: 0.8382 - binary_accuracy: 0.8382 - val_loss: 0.4440 - val_acc: 0.8344 - val_binary_accuracy: 0.8344\n",
      "Epoch 12/50\n",
      "4s - loss: 0.4239 - acc: 0.8414 - binary_accuracy: 0.8414 - val_loss: 0.4243 - val_acc: 0.8380 - val_binary_accuracy: 0.8380\n",
      "Epoch 13/50\n",
      "6s - loss: 0.4049 - acc: 0.8459 - binary_accuracy: 0.8459 - val_loss: 0.4084 - val_acc: 0.8416 - val_binary_accuracy: 0.8416\n",
      "Epoch 14/50\n",
      "5s - loss: 0.3893 - acc: 0.8484 - binary_accuracy: 0.8484 - val_loss: 0.3949 - val_acc: 0.8458 - val_binary_accuracy: 0.8458\n",
      "Epoch 15/50\n",
      "5s - loss: 0.3764 - acc: 0.8531 - binary_accuracy: 0.8531 - val_loss: 0.3868 - val_acc: 0.8416 - val_binary_accuracy: 0.8416\n",
      "Epoch 16/50\n",
      "5s - loss: 0.3657 - acc: 0.8542 - binary_accuracy: 0.8542 - val_loss: 0.3776 - val_acc: 0.8438 - val_binary_accuracy: 0.8438\n",
      "Epoch 17/50\n",
      "4s - loss: 0.3567 - acc: 0.8573 - binary_accuracy: 0.8573 - val_loss: 0.3702 - val_acc: 0.8456 - val_binary_accuracy: 0.8456\n",
      "Epoch 18/50\n",
      "4s - loss: 0.3491 - acc: 0.8592 - binary_accuracy: 0.8592 - val_loss: 0.3618 - val_acc: 0.8514 - val_binary_accuracy: 0.8514\n",
      "Epoch 19/50\n",
      "6s - loss: 0.3424 - acc: 0.8606 - binary_accuracy: 0.8606 - val_loss: 0.3568 - val_acc: 0.8518 - val_binary_accuracy: 0.8518\n",
      "Epoch 20/50\n",
      "5s - loss: 0.3370 - acc: 0.8622 - binary_accuracy: 0.8622 - val_loss: 0.3546 - val_acc: 0.8510 - val_binary_accuracy: 0.8510\n",
      "Epoch 21/50\n",
      "5s - loss: 0.3324 - acc: 0.8625 - binary_accuracy: 0.8625 - val_loss: 0.3481 - val_acc: 0.8534 - val_binary_accuracy: 0.8534\n",
      "Epoch 22/50\n",
      "5s - loss: 0.3280 - acc: 0.8642 - binary_accuracy: 0.8642 - val_loss: 0.3448 - val_acc: 0.8538 - val_binary_accuracy: 0.8538\n",
      "Epoch 23/50\n",
      "5s - loss: 0.3241 - acc: 0.8656 - binary_accuracy: 0.8656 - val_loss: 0.3420 - val_acc: 0.8546 - val_binary_accuracy: 0.8546\n",
      "Epoch 24/50\n",
      "5s - loss: 0.3207 - acc: 0.8649 - binary_accuracy: 0.8649 - val_loss: 0.3403 - val_acc: 0.8554 - val_binary_accuracy: 0.8554\n",
      "Epoch 25/50\n",
      "5s - loss: 0.3180 - acc: 0.8669 - binary_accuracy: 0.8669 - val_loss: 0.3374 - val_acc: 0.8558 - val_binary_accuracy: 0.8558\n",
      "Epoch 26/50\n",
      "5s - loss: 0.3153 - acc: 0.8685 - binary_accuracy: 0.8685 - val_loss: 0.3372 - val_acc: 0.8564 - val_binary_accuracy: 0.8564\n",
      "Epoch 27/50\n",
      "5s - loss: 0.3130 - acc: 0.8695 - binary_accuracy: 0.8695 - val_loss: 0.3339 - val_acc: 0.8572 - val_binary_accuracy: 0.8572\n",
      "Epoch 28/50\n",
      "5s - loss: 0.3111 - acc: 0.8701 - binary_accuracy: 0.8701 - val_loss: 0.3326 - val_acc: 0.8584 - val_binary_accuracy: 0.8584\n",
      "Epoch 29/50\n",
      "5s - loss: 0.3088 - acc: 0.8701 - binary_accuracy: 0.8701 - val_loss: 0.3323 - val_acc: 0.8576 - val_binary_accuracy: 0.8576\n",
      "Epoch 30/50\n",
      "5s - loss: 0.3072 - acc: 0.8713 - binary_accuracy: 0.8713 - val_loss: 0.3312 - val_acc: 0.8582 - val_binary_accuracy: 0.8582\n",
      "Epoch 31/50\n",
      "5s - loss: 0.3058 - acc: 0.8711 - binary_accuracy: 0.8711 - val_loss: 0.3296 - val_acc: 0.8594 - val_binary_accuracy: 0.8594\n",
      "Epoch 32/50\n",
      "5s - loss: 0.3042 - acc: 0.8717 - binary_accuracy: 0.8717 - val_loss: 0.3286 - val_acc: 0.8606 - val_binary_accuracy: 0.8606\n",
      "Epoch 33/50\n",
      "4s - loss: 0.3030 - acc: 0.8726 - binary_accuracy: 0.8726 - val_loss: 0.3281 - val_acc: 0.8588 - val_binary_accuracy: 0.8588\n",
      "Epoch 34/50\n",
      "5s - loss: 0.3017 - acc: 0.8728 - binary_accuracy: 0.8728 - val_loss: 0.3276 - val_acc: 0.8600 - val_binary_accuracy: 0.8600\n",
      "Epoch 35/50\n",
      "5s - loss: 0.3004 - acc: 0.8740 - binary_accuracy: 0.8740 - val_loss: 0.3268 - val_acc: 0.8594 - val_binary_accuracy: 0.8594\n",
      "Epoch 36/50\n",
      "5s - loss: 0.2996 - acc: 0.8731 - binary_accuracy: 0.8731 - val_loss: 0.3289 - val_acc: 0.8602 - val_binary_accuracy: 0.8602\n",
      "Epoch 37/50\n",
      "5s - loss: 0.2988 - acc: 0.8747 - binary_accuracy: 0.8747 - val_loss: 0.3278 - val_acc: 0.8616 - val_binary_accuracy: 0.8616\n",
      "Epoch 38/50\n",
      "5s - loss: 0.2978 - acc: 0.8745 - binary_accuracy: 0.8745 - val_loss: 0.3263 - val_acc: 0.8618 - val_binary_accuracy: 0.8618\n",
      "Epoch 39/50\n",
      "5s - loss: 0.2969 - acc: 0.8740 - binary_accuracy: 0.8740 - val_loss: 0.3283 - val_acc: 0.8608 - val_binary_accuracy: 0.8608\n",
      "Epoch 40/50\n",
      "5s - loss: 0.2960 - acc: 0.8749 - binary_accuracy: 0.8749 - val_loss: 0.3253 - val_acc: 0.8598 - val_binary_accuracy: 0.8598\n",
      "Epoch 41/50\n",
      "5s - loss: 0.2955 - acc: 0.8746 - binary_accuracy: 0.8746 - val_loss: 0.3252 - val_acc: 0.8602 - val_binary_accuracy: 0.8602\n",
      "Epoch 42/50\n",
      "5s - loss: 0.2950 - acc: 0.8746 - binary_accuracy: 0.8746 - val_loss: 0.3269 - val_acc: 0.8624 - val_binary_accuracy: 0.8624\n",
      "Epoch 43/50\n",
      "4s - loss: 0.2943 - acc: 0.8760 - binary_accuracy: 0.8760 - val_loss: 0.3250 - val_acc: 0.8600 - val_binary_accuracy: 0.8600\n",
      "Epoch 44/50\n",
      "5s - loss: 0.2936 - acc: 0.8759 - binary_accuracy: 0.8759 - val_loss: 0.3252 - val_acc: 0.8604 - val_binary_accuracy: 0.8604\n",
      "Epoch 45/50\n",
      "5s - loss: 0.2930 - acc: 0.8773 - binary_accuracy: 0.8773 - val_loss: 0.3263 - val_acc: 0.8620 - val_binary_accuracy: 0.8620\n",
      "Epoch 46/50\n",
      "5s - loss: 0.2927 - acc: 0.8761 - binary_accuracy: 0.8761 - val_loss: 0.3254 - val_acc: 0.8608 - val_binary_accuracy: 0.8608\n",
      "Epoch 47/50\n",
      "5s - loss: 0.2920 - acc: 0.8764 - binary_accuracy: 0.8764 - val_loss: 0.3294 - val_acc: 0.8580 - val_binary_accuracy: 0.8580\n",
      "Epoch 48/50\n",
      "5s - loss: 0.2916 - acc: 0.8772 - binary_accuracy: 0.8772 - val_loss: 0.3269 - val_acc: 0.8600 - val_binary_accuracy: 0.8600\n",
      "Epoch 49/50\n",
      "5s - loss: 0.2913 - acc: 0.8769 - binary_accuracy: 0.8769 - val_loss: 0.3250 - val_acc: 0.8598 - val_binary_accuracy: 0.8598\n",
      "Epoch 50/50\n",
      "5s - loss: 0.2908 - acc: 0.8772 - binary_accuracy: 0.8772 - val_loss: 0.3314 - val_acc: 0.8586 - val_binary_accuracy: 0.8586\n"
     ]
    }
   ],
   "source": [
    "model.compile(loss='binary_crossentropy', \n",
    "              optimizer='sgd', \n",
    "              metrics=['accuracy', 'binary_accuracy'])\n",
    "\n",
    "history = model.fit(train_data_features.toarray(), train_labels, \n",
    "                    validation_split=0.2,\n",
    "                    verbose=2, epochs= 50)\n",
    "# Max accuracy at 25 epochs is 85.76"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train and validation accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f5cfaa00f90>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5cfab03610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train and validation loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f5cfaa19590>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJS/17WGRAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f5cfaa49190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [conda env:tensorflow]",
   "language": "python",
   "name": "conda-env-tensorflow-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}