07_classifier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 431,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Embedding, LSTM, SpatialDropout1D\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from sklearn.model_selection import train_test_split\n",
    "from keras.utils.np_utils import to_categorical\n",
    "from keras.callbacks import EarlyStopping\n",
    "\n",
    "import gensim\n",
    "from gensim.models.doc2vec import Doc2Vec, TaggedDocument\n",
    "from keras.layers import Flatten"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 432,
   "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>labels</th>\n",
       "      <th>body</th>\n",
       "      <th>parent</th>\n",
       "      <th>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>supporting</td>\n",
       "      <td>no question want thank post topic birth contro...</td>\n",
       "      <td>updated september scroll downhi lengthy post c...</td>\n",
       "      <td>31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>supporting</td>\n",
       "      <td>holy shit thank much someone recently turn sup...</td>\n",
       "      <td>updated september scroll downhi lengthy post c...</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       labels                                               body  \\\n",
       "0  supporting  no question want thank post topic birth contro...   \n",
       "1  supporting  holy shit thank much someone recently turn sup...   \n",
       "\n",
       "                                              parent  score  \n",
       "0  updated september scroll downhi lengthy post c...     31  \n",
       "1  updated september scroll downhi lengthy post c...      7  "
      ]
     },
     "execution_count": 432,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pd.read_csv('datasets/labelled_dataset.csv',index_col=0)\n",
    "data = data[['labels','body','parent','score']]\n",
    "data.head(2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Checking Label Distribution and Assigning Weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 433,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "commenting     614\n",
       "supporting     299\n",
       "denying        137\n",
       "questioning    105\n",
       "Name: labels, dtype: int64"
      ]
     },
     "execution_count": 433,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.labels.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 434,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2e3425f4278>"
      ]
     },
     "execution_count": 434,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "data.labels.value_counts().plot(kind='bar')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 435,
   "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>labels</th>\n",
       "      <th>body</th>\n",
       "      <th>parent</th>\n",
       "      <th>score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>no question want thank post topic birth contro...</td>\n",
       "      <td>updated september scroll downhi lengthy post c...</td>\n",
       "      <td>31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>holy shit thank much someone recently turn sup...</td>\n",
       "      <td>updated september scroll downhi lengthy post c...</td>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   labels                                               body  \\\n",
       "0       0  no question want thank post topic birth contro...   \n",
       "1       0  holy shit thank much someone recently turn sup...   \n",
       "\n",
       "                                              parent  score  \n",
       "0  updated september scroll downhi lengthy post c...     31  \n",
       "1  updated september scroll downhi lengthy post c...      7  "
      ]
     },
     "execution_count": 435,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# turning into integers\n",
    "data.loc[data['labels'] == 'supporting', 'labels'] = 0\n",
    "data.loc[data['labels'] == 'denying', 'labels'] = 1\n",
    "data.loc[data['labels'] == 'questioning', 'labels'] = 2\n",
    "data.loc[data['labels'] == 'commenting', 'labels'] = 3\n",
    "data.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 436,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[1., 0., 0., 0.],\n",
       "       [1., 0., 0., 0.],\n",
       "       [0., 0., 0., 1.],\n",
       "       ...,\n",
       "       [0., 0., 0., 1.],\n",
       "       [0., 0., 0., 1.],\n",
       "       [0., 0., 0., 1.]], dtype=float32)"
      ]
     },
     "execution_count": 436,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# one-hot encoding\n",
    "labels = to_categorical(data['labels'], num_classes=4)\n",
    "labels"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Vectorizing with Doc2Vec"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 458,
   "metadata": {},
   "outputs": [],
   "source": [
    "from gensim.test.utils import common_texts\n",
    "from gensim.models.doc2vec import Doc2Vec, TaggedDocument\n",
    "import numpy as np\n",
    "\n",
    "# documents = [TaggedDocument(doc, [i]) for i, doc in enumerate(common_texts)]\n",
    "comments = [TaggedDocument(row[1][2], [index]) for index, row in data.iterrows()]\n",
    "source = [TaggedDocument(row[1][3], [index]) for index, row in data.iterrows()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 459,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W0722 11:04:36.337714 11096 base_any2vec.py:723] consider setting layer size to a multiple of 4 for greater performance\n",
      "W0722 11:04:36.338711 11096 doc2vec.py:1314] Each 'words' should be a list of words (usually unicode strings). First 'words' here is instead plain <class 'str'>.\n",
      "W0722 11:04:36.485318 11096 base_any2vec.py:1386] under 10 jobs per worker: consider setting a smaller `batch_words' for smoother alpha decay\n",
      "W0722 11:04:36.487313 11096 base_any2vec.py:723] consider setting layer size to a multiple of 4 for greater performance\n",
      "W0722 11:04:36.488310 11096 doc2vec.py:1314] Each 'words' should be a list of words (usually unicode strings). First 'words' here is instead plain <class 'str'>.\n",
      "W0722 11:04:37.018743 11096 base_any2vec.py:1386] under 10 jobs per worker: consider setting a smaller `batch_words' for smoother alpha decay\n"
     ]
    }
   ],
   "source": [
    "model_comments = Doc2Vec(comments, vector_size=5, window=2, min_count=1, workers=4)\n",
    "model_source = Doc2Vec(source, vector_size=5, window=2, min_count=1, workers=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 497,
   "metadata": {},
   "outputs": [],
   "source": [
    "final_vec_list = [\n",
    "    np.append(model_comments.docvecs[i], model_source.docvecs[i])\n",
    "    for i in range(len(model_comments.docvecs))\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 498,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_vec_list = np.asarray(final_vec_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1042,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, Y_train, Y_test = train_test_split(np_vec_list, labels, test_size=0.2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling with LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1043,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1155, 4)"
      ]
     },
     "execution_count": 1043,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1044,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(924, 10)"
      ]
     },
     "execution_count": 1044,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1045,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_train = np.asarray(Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1046,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_test = np.asarray(Y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1047,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = np.reshape(X_train, (X_train.shape[0], 1, X_train.shape[1]))\n",
    "X_test = np.reshape(X_test, (X_test.shape[0], 1, X_test.shape[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1049,
   "metadata": {},
   "outputs": [],
   "source": [
    "# label weights\n",
    "\n",
    "c = 12.2\n",
    "s = 22.9\n",
    "d = 40.1\n",
    "q = 50.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1050,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\users\\limri\\appdata\\local\\programs\\python\\python37\\lib\\site-packages\\ipykernel_launcher.py:7: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(activation=\"softmax\", units=4)`\n",
      "  import sys\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 739 samples, validate on 185 samples\n",
      "Epoch 1/100\n",
      "739/739 [==============================] - 13s 18ms/step - loss: 29.1703 - acc: 0.3694 - val_loss: 30.4311 - val_acc: 0.4649\n",
      "Epoch 2/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 29.1141 - acc: 0.4723 - val_loss: 30.3913 - val_acc: 0.4919\n",
      "Epoch 3/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 29.0538 - acc: 0.5304 - val_loss: 30.3627 - val_acc: 0.5081\n",
      "Epoch 4/100\n",
      "739/739 [==============================] - 0s 61us/step - loss: 29.0046 - acc: 0.5521 - val_loss: 30.3335 - val_acc: 0.5135\n",
      "Epoch 5/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.9638 - acc: 0.5535 - val_loss: 30.3077 - val_acc: 0.5135\n",
      "Epoch 6/100\n",
      "739/739 [==============================] - 0s 66us/step - loss: 28.9180 - acc: 0.5535 - val_loss: 30.2903 - val_acc: 0.5135\n",
      "Epoch 7/100\n",
      "739/739 [==============================] - 0s 72us/step - loss: 28.8817 - acc: 0.5535 - val_loss: 30.2730 - val_acc: 0.5135\n",
      "Epoch 8/100\n",
      "739/739 [==============================] - 0s 66us/step - loss: 28.8472 - acc: 0.5535 - val_loss: 30.2555 - val_acc: 0.5135\n",
      "Epoch 9/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.8160 - acc: 0.5535 - val_loss: 30.2475 - val_acc: 0.5135\n",
      "Epoch 10/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.7862 - acc: 0.5535 - val_loss: 30.2435 - val_acc: 0.5135\n",
      "Epoch 11/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.7666 - acc: 0.5535 - val_loss: 30.2388 - val_acc: 0.5135\n",
      "Epoch 12/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.7433 - acc: 0.5535 - val_loss: 30.2396 - val_acc: 0.5135\n",
      "Epoch 13/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.7321 - acc: 0.5535 - val_loss: 30.2397 - val_acc: 0.5135\n",
      "Epoch 14/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.7143 - acc: 0.5535 - val_loss: 30.2449 - val_acc: 0.5135\n",
      "Epoch 15/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.7095 - acc: 0.5535 - val_loss: 30.2484 - val_acc: 0.5135\n",
      "Epoch 16/100\n",
      "739/739 [==============================] - 0s 80us/step - loss: 28.6924 - acc: 0.5535 - val_loss: 30.2555 - val_acc: 0.5135\n",
      "Epoch 17/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.6872 - acc: 0.5535 - val_loss: 30.2636 - val_acc: 0.5135\n",
      "Epoch 18/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.6750 - acc: 0.5535 - val_loss: 30.2666 - val_acc: 0.5135\n",
      "Epoch 19/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6705 - acc: 0.5535 - val_loss: 30.2732 - val_acc: 0.5135\n",
      "Epoch 20/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.6602 - acc: 0.5535 - val_loss: 30.2765 - val_acc: 0.5135\n",
      "Epoch 21/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6559 - acc: 0.5535 - val_loss: 30.2794 - val_acc: 0.5135\n",
      "Epoch 22/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.6546 - acc: 0.5535 - val_loss: 30.2732 - val_acc: 0.5135\n",
      "Epoch 23/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6428 - acc: 0.5535 - val_loss: 30.2836 - val_acc: 0.5135\n",
      "Epoch 24/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6392 - acc: 0.5535 - val_loss: 30.2869 - val_acc: 0.5135\n",
      "Epoch 25/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.6323 - acc: 0.5535 - val_loss: 30.2802 - val_acc: 0.5135\n",
      "Epoch 26/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.6277 - acc: 0.5535 - val_loss: 30.2874 - val_acc: 0.5135\n",
      "Epoch 27/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.6228 - acc: 0.5535 - val_loss: 30.2870 - val_acc: 0.5135\n",
      "Epoch 28/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6188 - acc: 0.5535 - val_loss: 30.2944 - val_acc: 0.5135\n",
      "Epoch 29/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.6133 - acc: 0.5535 - val_loss: 30.2883 - val_acc: 0.5135\n",
      "Epoch 30/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.6040 - acc: 0.5535 - val_loss: 30.2875 - val_acc: 0.5135\n",
      "Epoch 31/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.6007 - acc: 0.5535 - val_loss: 30.2883 - val_acc: 0.5135\n",
      "Epoch 32/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.5934 - acc: 0.5535 - val_loss: 30.2950 - val_acc: 0.5135\n",
      "Epoch 33/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.5895 - acc: 0.5535 - val_loss: 30.2971 - val_acc: 0.5135\n",
      "Epoch 34/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.5842 - acc: 0.5535 - val_loss: 30.2995 - val_acc: 0.5135\n",
      "Epoch 35/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.5769 - acc: 0.5535 - val_loss: 30.2935 - val_acc: 0.5135\n",
      "Epoch 36/100\n",
      "739/739 [==============================] - 0s 76us/step - loss: 28.5735 - acc: 0.5535 - val_loss: 30.2987 - val_acc: 0.5135\n",
      "Epoch 37/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.5650 - acc: 0.5535 - val_loss: 30.3030 - val_acc: 0.5135\n",
      "Epoch 38/100\n",
      "739/739 [==============================] - 0s 67us/step - loss: 28.5614 - acc: 0.5535 - val_loss: 30.3079 - val_acc: 0.5135\n",
      "Epoch 39/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.5548 - acc: 0.5535 - val_loss: 30.3167 - val_acc: 0.5135\n",
      "Epoch 40/100\n",
      "739/739 [==============================] - 0s 67us/step - loss: 28.5483 - acc: 0.5535 - val_loss: 30.3133 - val_acc: 0.5135\n",
      "Epoch 41/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.5420 - acc: 0.5535 - val_loss: 30.3150 - val_acc: 0.5135\n",
      "Epoch 42/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.5394 - acc: 0.5535 - val_loss: 30.3237 - val_acc: 0.5135\n",
      "Epoch 43/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.5319 - acc: 0.5535 - val_loss: 30.3188 - val_acc: 0.5135\n",
      "Epoch 44/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.5254 - acc: 0.5535 - val_loss: 30.3211 - val_acc: 0.5135\n",
      "Epoch 45/100\n",
      "739/739 [==============================] - 0s 61us/step - loss: 28.5194 - acc: 0.5535 - val_loss: 30.3207 - val_acc: 0.5135\n",
      "Epoch 46/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.5140 - acc: 0.5548 - val_loss: 30.3198 - val_acc: 0.5135\n",
      "Epoch 47/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.5093 - acc: 0.5535 - val_loss: 30.3233 - val_acc: 0.5081\n",
      "Epoch 48/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.5048 - acc: 0.5521 - val_loss: 30.3330 - val_acc: 0.5081\n",
      "Epoch 49/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.4970 - acc: 0.5521 - val_loss: 30.3296 - val_acc: 0.5027\n",
      "Epoch 50/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.4959 - acc: 0.5507 - val_loss: 30.3417 - val_acc: 0.5027\n",
      "Epoch 51/100\n",
      "739/739 [==============================] - 0s 67us/step - loss: 28.4913 - acc: 0.5494 - val_loss: 30.3249 - val_acc: 0.5027\n",
      "Epoch 52/100\n",
      "739/739 [==============================] - 0s 81us/step - loss: 28.4821 - acc: 0.5494 - val_loss: 30.3331 - val_acc: 0.5027\n",
      "Epoch 53/100\n",
      "739/739 [==============================] - 0s 80us/step - loss: 28.4763 - acc: 0.5494 - val_loss: 30.3386 - val_acc: 0.5027\n",
      "Epoch 54/100\n",
      "739/739 [==============================] - 0s 70us/step - loss: 28.4717 - acc: 0.5507 - val_loss: 30.3441 - val_acc: 0.5027\n",
      "Epoch 55/100\n",
      "739/739 [==============================] - 0s 76us/step - loss: 28.4660 - acc: 0.5494 - val_loss: 30.3489 - val_acc: 0.5027\n",
      "Epoch 56/100\n",
      "739/739 [==============================] - 0s 78us/step - loss: 28.4646 - acc: 0.5480 - val_loss: 30.3509 - val_acc: 0.4973\n",
      "Epoch 57/100\n",
      "739/739 [==============================] - 0s 66us/step - loss: 28.4590 - acc: 0.5494 - val_loss: 30.3697 - val_acc: 0.5027\n",
      "Epoch 58/100\n",
      "739/739 [==============================] - 0s 64us/step - loss: 28.4516 - acc: 0.5480 - val_loss: 30.3673 - val_acc: 0.4919\n",
      "Epoch 59/100\n",
      "739/739 [==============================] - 0s 76us/step - loss: 28.4483 - acc: 0.5386 - val_loss: 30.3707 - val_acc: 0.4865\n",
      "Epoch 60/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.4419 - acc: 0.5426 - val_loss: 30.3729 - val_acc: 0.4811\n",
      "Epoch 61/100\n",
      "739/739 [==============================] - 0s 66us/step - loss: 28.4424 - acc: 0.5440 - val_loss: 30.3824 - val_acc: 0.4811\n",
      "Epoch 62/100\n",
      "739/739 [==============================] - 0s 70us/step - loss: 28.4372 - acc: 0.5426 - val_loss: 30.3686 - val_acc: 0.4703\n",
      "Epoch 63/100\n",
      "739/739 [==============================] - 0s 66us/step - loss: 28.4359 - acc: 0.5440 - val_loss: 30.3736 - val_acc: 0.4703\n",
      "Epoch 64/100\n",
      "739/739 [==============================] - 0s 69us/step - loss: 28.4264 - acc: 0.5440 - val_loss: 30.3779 - val_acc: 0.4703\n",
      "Epoch 65/100\n",
      "739/739 [==============================] - 0s 86us/step - loss: 28.4232 - acc: 0.5386 - val_loss: 30.3787 - val_acc: 0.4595\n",
      "Epoch 66/100\n",
      "739/739 [==============================] - 0s 67us/step - loss: 28.4221 - acc: 0.5359 - val_loss: 30.3693 - val_acc: 0.4649\n",
      "Epoch 67/100\n",
      "739/739 [==============================] - 0s 63us/step - loss: 28.4161 - acc: 0.5345 - val_loss: 30.3732 - val_acc: 0.4649\n",
      "Epoch 68/100\n",
      "739/739 [==============================] - 0s 62us/step - loss: 28.4129 - acc: 0.5345 - val_loss: 30.3776 - val_acc: 0.4595\n",
      "Epoch 69/100\n",
      "739/739 [==============================] - 0s 73us/step - loss: 28.4103 - acc: 0.5345 - val_loss: 30.3851 - val_acc: 0.4595\n",
      "Epoch 70/100\n",
      "739/739 [==============================] - 0s 65us/step - loss: 28.4117 - acc: 0.5372 - val_loss: 30.4110 - val_acc: 0.4595\n",
      "Epoch 71/100\n",
      "739/739 [==============================] - 0s 179us/step - loss: 28.4041 - acc: 0.5372 - val_loss: 30.4090 - val_acc: 0.4595\n",
      "Epoch 72/100\n",
      "739/739 [==============================] - 0s 238us/step - loss: 28.3989 - acc: 0.5372 - val_loss: 30.4195 - val_acc: 0.4486\n",
      "Epoch 73/100\n",
      "739/739 [==============================] - 0s 157us/step - loss: 28.3994 - acc: 0.5372 - val_loss: 30.4281 - val_acc: 0.4486\n",
      "Epoch 74/100\n",
      "739/739 [==============================] - 0s 144us/step - loss: 28.3949 - acc: 0.5372 - val_loss: 30.4363 - val_acc: 0.4486\n",
      "Epoch 75/100\n",
      "739/739 [==============================] - 0s 118us/step - loss: 28.3956 - acc: 0.5372 - val_loss: 30.4408 - val_acc: 0.4541\n",
      "Epoch 76/100\n",
      "739/739 [==============================] - 0s 121us/step - loss: 28.3932 - acc: 0.5359 - val_loss: 30.4192 - val_acc: 0.4486\n",
      "Epoch 77/100\n",
      "739/739 [==============================] - 0s 156us/step - loss: 28.3872 - acc: 0.5359 - val_loss: 30.4333 - val_acc: 0.4486\n",
      "Epoch 78/100\n",
      "739/739 [==============================] - 0s 123us/step - loss: 28.3859 - acc: 0.5372 - val_loss: 30.4417 - val_acc: 0.4486\n",
      "Epoch 79/100\n",
      "739/739 [==============================] - 0s 91us/step - loss: 28.3815 - acc: 0.5359 - val_loss: 30.4444 - val_acc: 0.4486\n",
      "Epoch 80/100\n",
      "739/739 [==============================] - 0s 86us/step - loss: 28.3812 - acc: 0.5318 - val_loss: 30.4416 - val_acc: 0.4486\n",
      "Epoch 81/100\n",
      "739/739 [==============================] - 0s 81us/step - loss: 28.3785 - acc: 0.5291 - val_loss: 30.4428 - val_acc: 0.4486\n",
      "Epoch 82/100\n",
      "739/739 [==============================] - 0s 86us/step - loss: 28.3763 - acc: 0.5304 - val_loss: 30.4498 - val_acc: 0.4541\n",
      "Epoch 83/100\n",
      "739/739 [==============================] - 0s 109us/step - loss: 28.3774 - acc: 0.5304 - val_loss: 30.4608 - val_acc: 0.4541\n",
      "Epoch 84/100\n",
      "739/739 [==============================] - 0s 93us/step - loss: 28.3828 - acc: 0.5291 - val_loss: 30.4515 - val_acc: 0.4541\n",
      "Epoch 85/100\n",
      "739/739 [==============================] - 0s 82us/step - loss: 28.3725 - acc: 0.5332 - val_loss: 30.4476 - val_acc: 0.4541\n",
      "Epoch 86/100\n",
      "739/739 [==============================] - 0s 80us/step - loss: 28.3677 - acc: 0.5291 - val_loss: 30.4558 - val_acc: 0.4595\n",
      "Epoch 87/100\n",
      "739/739 [==============================] - 0s 80us/step - loss: 28.3673 - acc: 0.5277 - val_loss: 30.4619 - val_acc: 0.4595\n",
      "Epoch 88/100\n",
      "739/739 [==============================] - 0s 92us/step - loss: 28.3657 - acc: 0.5277 - val_loss: 30.4756 - val_acc: 0.4595\n",
      "Epoch 89/100\n",
      "739/739 [==============================] - 0s 213us/step - loss: 28.3664 - acc: 0.5291 - val_loss: 30.4762 - val_acc: 0.4541\n",
      "Epoch 90/100\n",
      "739/739 [==============================] - 0s 102us/step - loss: 28.3624 - acc: 0.5318 - val_loss: 30.4767 - val_acc: 0.4649\n",
      "Epoch 91/100\n",
      "739/739 [==============================] - 0s 83us/step - loss: 28.3616 - acc: 0.5250 - val_loss: 30.4805 - val_acc: 0.4649\n",
      "Epoch 92/100\n",
      "739/739 [==============================] - 0s 82us/step - loss: 28.3646 - acc: 0.5250 - val_loss: 30.4761 - val_acc: 0.4595\n",
      "Epoch 93/100\n",
      "739/739 [==============================] - 0s 87us/step - loss: 28.3582 - acc: 0.5250 - val_loss: 30.4855 - val_acc: 0.4595\n",
      "Epoch 94/100\n",
      "739/739 [==============================] - 0s 86us/step - loss: 28.3570 - acc: 0.5250 - val_loss: 30.4882 - val_acc: 0.4595\n",
      "Epoch 95/100\n",
      "739/739 [==============================] - 0s 84us/step - loss: 28.3571 - acc: 0.5237 - val_loss: 30.4973 - val_acc: 0.4595\n",
      "Epoch 96/100\n",
      "739/739 [==============================] - 0s 90us/step - loss: 28.3553 - acc: 0.5223 - val_loss: 30.5023 - val_acc: 0.4486\n",
      "Epoch 97/100\n",
      "739/739 [==============================] - 0s 84us/step - loss: 28.3597 - acc: 0.5223 - val_loss: 30.5058 - val_acc: 0.4486\n",
      "Epoch 98/100\n",
      "739/739 [==============================] - 0s 90us/step - loss: 28.3567 - acc: 0.5264 - val_loss: 30.5046 - val_acc: 0.4595\n",
      "Epoch 99/100\n",
      "739/739 [==============================] - 0s 118us/step - loss: 28.3510 - acc: 0.5250 - val_loss: 30.5065 - val_acc: 0.4486\n",
      "Epoch 100/100\n",
      "739/739 [==============================] - 0s 104us/step - loss: 28.3551 - acc: 0.5210 - val_loss: 30.4969 - val_acc: 0.4486\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_102 (LSTM)              (None, 30)                4920      \n",
      "_________________________________________________________________\n",
      "dense_95 (Dense)             (None, 4)                 124       \n",
      "=================================================================\n",
      "Total params: 5,044\n",
      "Trainable params: 5,044\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "epochs = 100\n",
    "emb_dim = 128\n",
    "batch_size = 64\n",
    "\n",
    "model = Sequential()\n",
    "model.add(LSTM(30))\n",
    "model.add(Dense(output_dim=4,activation='softmax'))\n",
    "model.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['acc'])\n",
    "# history = model.fit(X_train,Y_train,batch_size=batch_size,epochs=epochs,class_weight={0:s,1:d,2:q,3:c},validation_split=0.2,callbacks=[EarlyStopping(monitor='val_loss',patience=7, min_delta=0.0001)])\n",
    "history = model.fit(X_train,Y_train,batch_size=batch_size,epochs=epochs,class_weight={0:s,1:d,2:q,3:c},validation_split=0.2)\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1051,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "231/231 [==============================] - 0s 147us/step\n",
      "Test set\n",
      "  Loss: 1.307\n",
      "  Accuracy: 0.411\n"
     ]
    }
   ],
   "source": [
    "accr = model.evaluate(X_test,Y_test)\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1052,
   "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"
    },
    {
     "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": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "acc = history.history['acc']\n",
    "val_acc = history.history['val_acc']\n",
    "loss = history.history['loss']\n",
    "val_loss = history.history['val_loss']\n",
    "\n",
    "epochs = range(1, len(acc) + 1)\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1053,
   "metadata": {},
   "outputs": [],
   "source": [
    "np_vec_shaped = np.reshape(np_vec_list,(np_vec_list.shape[0],1,np_vec_list.shape[1]))\n",
    "pred = model.predict(np_vec_shaped)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1054,
   "metadata": {},
   "outputs": [],
   "source": [
    "ary = []\n",
    "lb = ['supporting', 'denying', 'questioning', 'commenting']\n",
    "for i in range(pred.shape[0]):\n",
    "    ary.append(lb[np.argmax(pred[i])])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1055,
   "metadata": {},
   "outputs": [],
   "source": [
    "sup = [1.,0.,0.,0.]\n",
    "den = [0.,1.,0.,0.]\n",
    "que = [0.,0.,1.,0.]\n",
    "com = [0.,0.,0.,1.]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1056,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = []\n",
    "for i in range(np_vec_shaped.shape[0]):\n",
    "    if( list(np_vec_shaped[i]) == sup ):\n",
    "        test.append('supporting')\n",
    "    elif( list(np_vec_shaped[i]) == den ):\n",
    "        test.append('denying')\n",
    "    elif( list(np_vec_shaped[i]) == que ):\n",
    "        test.append('questioning')\n",
    "    else:\n",
    "        test.append('commenting')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1057,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = pd.DataFrame(test)\n",
    "results.columns = ['test']\n",
    "results['predictions'] = ary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1058,
   "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>test</th>\n",
       "      <th>predictions</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>commenting</td>\n",
       "      <td>commenting</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>commenting</td>\n",
       "      <td>commenting</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>commenting</td>\n",
       "      <td>commenting</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>commenting</td>\n",
       "      <td>commenting</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>commenting</td>\n",
       "      <td>commenting</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         test predictions\n",
       "0  commenting  commenting\n",
       "1  commenting  commenting\n",
       "2  commenting  commenting\n",
       "3  commenting  commenting\n",
       "4  commenting  commenting"
      ]
     },
     "execution_count": 1058,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "results.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1059,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2e3b81f97f0>"
      ]
     },
     "execution_count": 1059,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "results.predictions.value_counts().plot(kind='bar')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1060,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1155/1155 [==============================] - 0s 113us/step\n",
      "Test set\n",
      "  Loss: 1.264\n",
      "  Accuracy: 0.488\n"
     ]
    }
   ],
   "source": [
    "accr = model.evaluate(np_vec_shaped,labels)\n",
    "print('Test set\\n  Loss: {:0.3f}\\n  Accuracy: {:0.3f}'.format(accr[0],accr[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.7.0b4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}