4.0 Ad classification and Likes prediction with LSTM models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 102,
   "id": "bbac8b59",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import re\n",
    "import jieba as jb\n",
    "import tensorflow as tf\n",
    "from sklearn.model_selection import train_test_split\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "#from text_cleaner import remove\n",
    "#from text_cleaner.processor.chinese import CHINESE, CHINESE_SYMBOLS_AND_PUNCTUATION\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "a8b946b2",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(r'/Users/cairo/Google Drive/wechat data/ad data/AntzbBackData10k.csv', sep=',', error_bad_lines=False, index_col=False, dtype='unicode')\n",
    "\n",
    "df=df[[\"account\", 'advertisement','title', \"content\", \"clicksCount\", \"orderNum\", \"originalFlag\", \"likeCount\", \"publicTime\"]]\n",
    "\n",
    "df[df.isnull().values==True]\n",
    "df = df[pd.notnull(df['title'])]\n",
    "df = df[pd.notnull(df['content'])]\n",
    "df = df[pd.notnull(df['advertisement'])]\n",
    "\n",
    "df[\"publicTime\"] = pd.to_datetime(df[\"publicTime\"])\n",
    "#df[\"publicTime\"] = dd[\"publicTime\"].dt.to_period('D')\n",
    "df[\"clicksCount\"] = df[\"clicksCount\"].astype(float)\n",
    "df[\"likeCount\"] = df[\"likeCount\"].astype(float)\n",
    "df[\"orderNum\"] = df[\"orderNum\"].astype(float)\n",
    "df[\"originalFlag\"] = df[\"originalFlag\"].astype(float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "e1f018c5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "944"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['account'].nunique() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "b41d40cc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "C    6846\n",
       "A    1935\n",
       "B    1214\n",
       "Name: advertisement, dtype: int64"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['advertisement'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "14621b76",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Text preprocessing\n",
    "\n",
    "rule = re.compile(u\"[^a-zA-Z0-9\\u4E00-\\u9FA5]\")\n",
    "def remove_punctuation(line):\n",
    "    line = rule.sub('',str(line))\n",
    "    return line\n",
    "\n",
    "\n",
    "def stopwordslist(filepath):  \n",
    "    stopwords = [line.strip() for line in open(filepath, 'r', encoding='utf-8').readlines()]  \n",
    "    return stopwords  \n",
    " \n",
    "#load the stopword file\n",
    "stopwords = stopwordslist(\"/Users/cairo/Google Drive/wechat data/stopword.txt\")\n",
    "\n",
    "\n",
    "\n",
    "line = str(df[\"content\"][0])\n",
    "line\n",
    "\n",
    "line.strip()\n",
    "\n",
    "line.strip()==''\n",
    "\n",
    "rule = re.compile(u\"[^a-zA-Z0-9\\u4E00-\\u9FA5]\")\n",
    "\n",
    "line2 = rule.sub('',line)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "e9d443b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Clean title and content of the articles separately\n",
    "\n",
    "df['clean_title'] = df['title'].map(lambda x: re.sub('\\W+','', x))\n",
    "\n",
    "df['clean_content'] = df['content'].map(lambda x: re.sub('\\W+','', x))\n",
    "\n",
    "\n",
    "df['cut_title'] = df['clean_title'].apply(lambda x: \" \".join([w for w in list(jb.cut(x)) if w not in stopwords]))\n",
    "\n",
    "df['cut_content'] = df['clean_content'].apply(lambda x: \" \".join([w for w in list(jb.cut(x)) if w not in stopwords]))\n",
    "\n",
    "\n",
    "#The combined text emphasized on the text in title by repeating it three times\n",
    "df['combined_text'] = df['cut_title'] + df['cut_title'] + df['cut_title'] + df['cut_content']\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "97fad3b3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There are 23871 unique words\n",
      "There are 310898 unique words\n",
      "(9997, 550)\n",
      "(9997, 3)\n"
     ]
    }
   ],
   "source": [
    "\n",
    "MAX_NB_WORDS = 50000\n",
    "# set max sequence length\n",
    "MAX_SEQUENCE_LENGTH_title = 50\n",
    "MAX_SEQUENCE_LENGTH_content = 500\n",
    "# set Embeddingceng dimensions\n",
    "EMBEDDING_DIM = 100 \n",
    "tokenizer = tf.keras.preprocessing.text.Tokenizer(num_words=MAX_NB_WORDS, filters='!\"#$%&()*+,-./:;<=>?@[\\]^_`{|}~', lower=True)\n",
    "\n",
    "tokenizer.fit_on_texts(df['cut_title'].values)\n",
    "word_index = tokenizer.word_index\n",
    "print('There are %s unique words' % len(word_index))\n",
    "X = tokenizer.texts_to_sequences(df['cut_title'].values)\n",
    "X = tf.keras.preprocessing.sequence.pad_sequences(X, maxlen=MAX_SEQUENCE_LENGTH_title) #padding \n",
    "\n",
    "\n",
    "tokenizer.fit_on_texts(df['cut_content'].values)\n",
    "word_index = tokenizer.word_index\n",
    "print('There are %s unique words' % len(word_index))\n",
    "X = tokenizer.texts_to_sequences(df['cut_content'].values)\n",
    "X = tf.keras.preprocessing.sequence.pad_sequences(X, maxlen=MAX_SEQUENCE_LENGTH_content) \n",
    "\n",
    "\n",
    "tokenizer.fit_on_texts(df['combined_text'].values)\n",
    "word_index = tokenizer.word_index\n",
    "X = tokenizer.texts_to_sequences(df['combined_text'].values)\n",
    "X = tf.keras.preprocessing.sequence.pad_sequences(X, maxlen=MAX_SEQUENCE_LENGTH_title + MAX_SEQUENCE_LENGTH_content)\n",
    "\n",
    " \n",
    "#onehot coding the dependent variable\n",
    "Y = pd.get_dummies(df['advertisement']).values\n",
    " \n",
    "Y2 = df['likeCount'].values\n",
    "\n",
    "print(X.shape)\n",
    "print(Y.shape)\n",
    "\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X,Y, test_size = 0.20, random_state = 42)\n",
    "#print(X_train.shape,Y_train.shape)\n",
    "#print(X_test.shape,Y_test.shape)\n",
    "\n",
    "X_train, X_test, Y2_train, Y2_test = train_test_split(X,Y2, test_size = 0.20, random_state = 42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "47781b27",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_3 (Embedding)      (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "lstm_4 (LSTM)                (None, 32)                17024     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 3)                 99        \n",
      "=================================================================\n",
      "Total params: 5,017,123\n",
      "Trainable params: 5,017,123\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_4 (Embedding)      (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "bidirectional_1 (Bidirection (None, 200)               160800    \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 3)                 603       \n",
      "=================================================================\n",
      "Total params: 5,161,403\n",
      "Trainable params: 5,161,403\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Model: \"sequential_5\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_5 (Embedding)      (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "lstm_6 (LSTM)                (None, 550, 32)           17024     \n",
      "_________________________________________________________________\n",
      "lstm_7 (LSTM)                (None, 32)                8320      \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 3)                 99        \n",
      "=================================================================\n",
      "Total params: 5,025,443\n",
      "Trainable params: 5,025,443\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "######## Models for ad classifcation \n",
    "#vanila LSTM \n",
    "model1 = tf.keras.Sequential()\n",
    "model1.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model1.add(tf.keras.layers.LSTM(32, dropout=0.2, recurrent_dropout=0.2))  \n",
    "model1.add(tf.keras.layers.Dense(3, activation='softmax'))\n",
    "model1.compile(loss='categorical_crossentropy', optimizer='adam', metrics=[tf.keras.metrics.CategoricalAccuracy(), tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])\n",
    "print(model1.summary())\n",
    "\n",
    "#bidirectional LSTM \n",
    "model2 = tf.keras.Sequential()\n",
    "model2.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model2.add(tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(100, dropout=0.2, recurrent_dropout=0.2)))\n",
    "model2.add(tf.keras.layers.Dense(3, activation='softmax'))\n",
    "model2.compile(loss='categorical_crossentropy', optimizer='adam', metrics=[tf.keras.metrics.CategoricalAccuracy(), tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])\n",
    "print(model2.summary())\n",
    "\n",
    "\n",
    "#stack two LSTM together\n",
    "model3 = tf.keras.Sequential()\n",
    "model3.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model3.add(tf.keras.layers.LSTM(32, dropout=0.2, recurrent_dropout=0.2, return_sequences=True))  \n",
    "model3.add(tf.keras.layers.LSTM(32))\n",
    "model3.add(tf.keras.layers.Dense(3, activation='softmax'))\n",
    "model3.compile(loss='categorical_crossentropy', optimizer='adam', metrics=[tf.keras.metrics.CategoricalAccuracy(), tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])\n",
    "print(model3.summary())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "a6e333bd",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "125/125 [==============================] - 90s 716ms/step - loss: 0.0470 - categorical_accuracy: 0.9845 - precision_3: 0.9863 - recall_3: 0.9840 - val_loss: 0.7679 - val_categorical_accuracy: 0.7970 - val_precision_3: 0.8053 - val_recall_3: 0.7908\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 105s 830ms/step - loss: 0.0354 - categorical_accuracy: 0.9886 - precision_3: 0.9897 - recall_3: 0.9871 - val_loss: 0.7953 - val_categorical_accuracy: 0.8095 - val_precision_3: 0.8114 - val_recall_3: 0.8033\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 64s 508ms/step - loss: 0.0273 - categorical_accuracy: 0.9914 - precision_3: 0.9921 - recall_3: 0.9909 - val_loss: 0.8015 - val_categorical_accuracy: 0.8005 - val_precision_3: 0.8079 - val_recall_3: 0.7958\n",
      "Epoch 1/20\n",
      "125/125 [==============================] - 199s 2s/step - loss: 0.0194 - categorical_accuracy: 0.9939 - precision_4: 0.9940 - recall_4: 0.9935 - val_loss: 0.8965 - val_categorical_accuracy: 0.8060 - val_precision_4: 0.8086 - val_recall_4: 0.8013\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 206s 2s/step - loss: 0.0191 - categorical_accuracy: 0.9942 - precision_4: 0.9945 - recall_4: 0.9939 - val_loss: 0.8210 - val_categorical_accuracy: 0.8115 - val_precision_4: 0.8155 - val_recall_4: 0.8098\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 198s 2s/step - loss: 0.0177 - categorical_accuracy: 0.9945 - precision_4: 0.9950 - recall_4: 0.9944 - val_loss: 0.8207 - val_categorical_accuracy: 0.8295 - val_precision_4: 0.8364 - val_recall_4: 0.8238\n",
      "Epoch 4/20\n",
      "125/125 [==============================] - 200s 2s/step - loss: 0.0155 - categorical_accuracy: 0.9956 - precision_4: 0.9961 - recall_4: 0.9955 - val_loss: 0.8547 - val_categorical_accuracy: 0.8175 - val_precision_4: 0.8251 - val_recall_4: 0.8148\n",
      "Epoch 5/20\n",
      "125/125 [==============================] - 201s 2s/step - loss: 0.0118 - categorical_accuracy: 0.9967 - precision_4: 0.9969 - recall_4: 0.9965 - val_loss: 0.8480 - val_categorical_accuracy: 0.8025 - val_precision_4: 0.8055 - val_recall_4: 0.8023\n",
      "Epoch 1/20\n",
      "125/125 [==============================] - 129s 1s/step - loss: 0.0577 - categorical_accuracy: 0.9816 - precision_5: 0.9827 - recall_5: 0.9806 - val_loss: 0.8050 - val_categorical_accuracy: 0.7775 - val_precision_5: 0.7854 - val_recall_5: 0.7768\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 112s 893ms/step - loss: 0.0377 - categorical_accuracy: 0.9889 - precision_5: 0.9900 - recall_5: 0.9885 - val_loss: 0.8506 - val_categorical_accuracy: 0.7970 - val_precision_5: 0.7988 - val_recall_5: 0.7948\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 107s 856ms/step - loss: 0.0315 - categorical_accuracy: 0.9910 - precision_5: 0.9916 - recall_5: 0.9907 - val_loss: 0.8654 - val_categorical_accuracy: 0.7935 - val_precision_5: 0.7969 - val_recall_5: 0.7913\n"
     ]
    }
   ],
   "source": [
    "#########\n",
    "epochs = 20\n",
    "batch_size = 64\n",
    " \n",
    "history1 = model1.fit(X, Y, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n",
    "history2 = model2.fit(X, Y, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n",
    "history3 = model3.fit(X, Y, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n",
    "#history = model.fit(X_train, Y_train, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "a898f847",
   "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": [
    "\n",
    "plt.title('Accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.plot(history1.history['categorical_accuracy'], label='train_lstm', color = \"red\")\n",
    "plt.plot(history1.history['val_categorical_accuracy'], label='test_lstm', linestyle='--', color = \"red\")\n",
    "plt.plot(history2.history['categorical_accuracy'], label='train_bidirect', color = \"navy\")\n",
    "plt.plot(history2.history['val_categorical_accuracy'], label='test_bidirect', linestyle='--', color = \"navy\")\n",
    "plt.plot(history3.history['categorical_accuracy'], label='train_stack', color = \"gold\")\n",
    "plt.plot(history3.history['val_categorical_accuracy'], label='test_stack', linestyle='--', color = \"gold\")\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "3244a835",
   "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.title('Loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.plot(history1.history['loss'], label='train_lstm', color = \"red\")\n",
    "plt.plot(history1.history['val_loss'], label='test_lstm', linestyle='--', color = \"red\")\n",
    "plt.plot(history2.history['loss'], label='train_bilstm', color = \"navy\")\n",
    "plt.plot(history2.history['val_loss'], label='test_bilstm', linestyle='--', color = \"navy\")\n",
    "plt.plot(history3.history['loss'], label='train_stack', color = \"gold\")\n",
    "plt.plot(history3.history['val_loss'], label='test_stack', linestyle='--', color = \"gold\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "717f2877",
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "plt.title('Recall')\n",
    "plt.plot(history1.history['recall_6'], label='train')\n",
    "plt.plot(history1.history['val_recall_6'], label='test')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "plt.title('Precision')\n",
    "plt.plot(history1.history['precision_6'], label='train')\n",
    "plt.plot(history1.history['val_precision_6'], label='test')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "de898126",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_21\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_21 (Embedding)     (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "lstm_28 (LSTM)               (None, 32)                17024     \n",
      "_________________________________________________________________\n",
      "dense_21 (Dense)             (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 5,017,057\n",
      "Trainable params: 5,017,057\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Model: \"sequential_22\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_22 (Embedding)     (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "bidirectional_7 (Bidirection (None, 200)               160800    \n",
      "_________________________________________________________________\n",
      "dense_22 (Dense)             (None, 1)                 201       \n",
      "=================================================================\n",
      "Total params: 5,161,001\n",
      "Trainable params: 5,161,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Model: \"sequential_23\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_23 (Embedding)     (None, 550, 100)          5000000   \n",
      "_________________________________________________________________\n",
      "lstm_30 (LSTM)               (None, 550, 32)           17024     \n",
      "_________________________________________________________________\n",
      "lstm_31 (LSTM)               (None, 32)                8320      \n",
      "_________________________________________________________________\n",
      "dense_23 (Dense)             (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 5,025,377\n",
      "Trainable params: 5,025,377\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "######## Models for regression\n",
    "#vanila LSTM \n",
    "model1a = tf.keras.Sequential()\n",
    "model1a.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model1a.add(tf.keras.layers.LSTM(32, dropout=0.2, recurrent_dropout=0.2))  \n",
    "model1a.add(tf.keras.layers.Dense(1))\n",
    "model1a.compile(loss='mean_squared_error', optimizer='adam', metrics=[tf.keras.metrics.MeanAbsolutePercentageError(),tf.keras.metrics.MeanSquaredError()])\n",
    "print(model1a.summary())\n",
    "\n",
    "#bidirectional LSTM \n",
    "model2a = tf.keras.Sequential()\n",
    "model2a.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model2a.add(tf.keras.layers.Bidirectional(tf.keras.layers.LSTM(100, dropout=0.2, recurrent_dropout=0.2)))\n",
    "model2a.add(tf.keras.layers.Dense(1))\n",
    "model2a.compile(loss='mean_squared_error', optimizer='adam', metrics=[tf.keras.metrics.MeanAbsolutePercentageError(),tf.keras.metrics.MeanSquaredError()])\n",
    "print(model2a.summary())\n",
    "\n",
    "\n",
    "#stack two LSTM together\n",
    "model3a = tf.keras.Sequential()\n",
    "model3a.add(tf.keras.layers.Embedding(MAX_NB_WORDS, EMBEDDING_DIM, input_length=X.shape[1]))\n",
    "model3a.add(tf.keras.layers.LSTM(32, dropout=0.2, recurrent_dropout=0.2, return_sequences=True))  \n",
    "model3a.add(tf.keras.layers.LSTM(32))\n",
    "model3a.add(tf.keras.layers.Dense(1))\n",
    "model3a.compile(loss='mean_squared_error', optimizer='adam', metrics=[tf.keras.metrics.MeanAbsolutePercentageError(),tf.keras.metrics.MeanSquaredError()])\n",
    "print(model3a.summary())\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "cc7108e5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "125/125 [==============================] - 71s 516ms/step - loss: 995619.1875 - mean_absolute_percentage_error: 169764160.0000 - mean_squared_error: 995619.1875 - val_loss: 801982.9375 - val_mean_absolute_percentage_error: 251628896.0000 - val_mean_squared_error: 801982.9375\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 63s 507ms/step - loss: 990514.5000 - mean_absolute_percentage_error: 310614240.0000 - mean_squared_error: 990514.5000 - val_loss: 798229.4375 - val_mean_absolute_percentage_error: 348431776.0000 - val_mean_squared_error: 798229.4375\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 64s 508ms/step - loss: 987226.8750 - mean_absolute_percentage_error: 399142176.0000 - mean_squared_error: 987226.8750 - val_loss: 795191.6875 - val_mean_absolute_percentage_error: 428380192.0000 - val_mean_squared_error: 795191.6875\n",
      "Epoch 4/20\n",
      "125/125 [==============================] - 65s 520ms/step - loss: 984486.3125 - mean_absolute_percentage_error: 471008512.0000 - mean_squared_error: 984486.3125 - val_loss: 792516.9375 - val_mean_absolute_percentage_error: 499857568.0000 - val_mean_squared_error: 792516.9375\n",
      "Epoch 5/20\n",
      "125/125 [==============================] - 64s 511ms/step - loss: 982020.7500 - mean_absolute_percentage_error: 545662592.0000 - mean_squared_error: 982020.7500 - val_loss: 790093.1250 - val_mean_absolute_percentage_error: 565180992.0000 - val_mean_squared_error: 790093.1250\n",
      "Epoch 6/20\n",
      "125/125 [==============================] - 63s 504ms/step - loss: 979725.5000 - mean_absolute_percentage_error: 606210240.0000 - mean_squared_error: 979725.5000 - val_loss: 787803.3125 - val_mean_absolute_percentage_error: 624186368.0000 - val_mean_squared_error: 787803.3125\n",
      "Epoch 7/20\n",
      "125/125 [==============================] - 63s 506ms/step - loss: 977544.8125 - mean_absolute_percentage_error: 651780672.0000 - mean_squared_error: 977544.8125 - val_loss: 785619.5625 - val_mean_absolute_percentage_error: 662350912.0000 - val_mean_squared_error: 785619.5625\n",
      "Epoch 8/20\n",
      "125/125 [==============================] - 64s 511ms/step - loss: 975429.1250 - mean_absolute_percentage_error: 649470592.0000 - mean_squared_error: 975429.1250 - val_loss: 783645.4375 - val_mean_absolute_percentage_error: 689264320.0000 - val_mean_squared_error: 783645.4375\n",
      "Epoch 9/20\n",
      "125/125 [==============================] - 68s 547ms/step - loss: 973371.8125 - mean_absolute_percentage_error: 606619904.0000 - mean_squared_error: 973371.8125 - val_loss: 781896.5625 - val_mean_absolute_percentage_error: 698490560.0000 - val_mean_squared_error: 781896.5625\n",
      "Epoch 10/20\n",
      "125/125 [==============================] - 64s 512ms/step - loss: 971383.8750 - mean_absolute_percentage_error: 590750848.0000 - mean_squared_error: 971383.8750 - val_loss: 780264.5000 - val_mean_absolute_percentage_error: 717259456.0000 - val_mean_squared_error: 780264.5000\n",
      "Epoch 11/20\n",
      "125/125 [==============================] - 64s 509ms/step - loss: 969444.0000 - mean_absolute_percentage_error: 555163648.0000 - mean_squared_error: 969444.0000 - val_loss: 778809.9375 - val_mean_absolute_percentage_error: 735836672.0000 - val_mean_squared_error: 778809.9375\n",
      "Epoch 12/20\n",
      "125/125 [==============================] - 64s 508ms/step - loss: 967526.6875 - mean_absolute_percentage_error: 507710496.0000 - mean_squared_error: 967526.6875 - val_loss: 777480.6875 - val_mean_absolute_percentage_error: 730264896.0000 - val_mean_squared_error: 777480.6875\n",
      "Epoch 13/20\n",
      "125/125 [==============================] - 63s 505ms/step - loss: 965604.1875 - mean_absolute_percentage_error: 451874816.0000 - mean_squared_error: 965604.1875 - val_loss: 775941.5000 - val_mean_absolute_percentage_error: 748943872.0000 - val_mean_squared_error: 775941.5000\n",
      "Epoch 14/20\n",
      "125/125 [==============================] - 70s 564ms/step - loss: 963670.4375 - mean_absolute_percentage_error: 390604896.0000 - mean_squared_error: 963670.4375 - val_loss: 775086.1875 - val_mean_absolute_percentage_error: 583350784.0000 - val_mean_squared_error: 775086.1875\n",
      "Epoch 15/20\n",
      "125/125 [==============================] - 63s 503ms/step - loss: 961718.1875 - mean_absolute_percentage_error: 269009824.0000 - mean_squared_error: 961718.1875 - val_loss: 773633.5000 - val_mean_absolute_percentage_error: 531053760.0000 - val_mean_squared_error: 773633.5000\n",
      "Epoch 16/20\n",
      "125/125 [==============================] - 63s 502ms/step - loss: 959785.8750 - mean_absolute_percentage_error: 226921312.0000 - mean_squared_error: 959785.8750 - val_loss: 771949.9375 - val_mean_absolute_percentage_error: 626380288.0000 - val_mean_squared_error: 771949.9375\n",
      "Epoch 17/20\n",
      "125/125 [==============================] - 63s 503ms/step - loss: 957909.5625 - mean_absolute_percentage_error: 210120640.0000 - mean_squared_error: 957909.5625 - val_loss: 771199.0000 - val_mean_absolute_percentage_error: 517168192.0000 - val_mean_squared_error: 771199.0000\n",
      "Epoch 18/20\n",
      "125/125 [==============================] - 66s 526ms/step - loss: 956076.2500 - mean_absolute_percentage_error: 196785408.0000 - mean_squared_error: 956076.2500 - val_loss: 768835.0625 - val_mean_absolute_percentage_error: 701663936.0000 - val_mean_squared_error: 768835.0625\n",
      "Epoch 19/20\n",
      "125/125 [==============================] - 63s 506ms/step - loss: 954294.9375 - mean_absolute_percentage_error: 182204656.0000 - mean_squared_error: 954294.9375 - val_loss: 768174.4375 - val_mean_absolute_percentage_error: 502431328.0000 - val_mean_squared_error: 768174.4375\n",
      "Epoch 20/20\n",
      "125/125 [==============================] - 63s 502ms/step - loss: 952452.1250 - mean_absolute_percentage_error: 173208672.0000 - mean_squared_error: 952452.1250 - val_loss: 766807.0000 - val_mean_absolute_percentage_error: 478805376.0000 - val_mean_squared_error: 766807.0000\n",
      "Epoch 1/20\n",
      "125/125 [==============================] - 214s 2s/step - loss: 974180.3125 - mean_absolute_percentage_error: 804296256.0000 - mean_squared_error: 974180.3125 - val_loss: 766917.1875 - val_mean_absolute_percentage_error: 1262762112.0000 - val_mean_squared_error: 766917.1875\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 200s 2s/step - loss: 950672.9375 - mean_absolute_percentage_error: 1574698880.0000 - mean_squared_error: 950672.9375 - val_loss: 751780.2500 - val_mean_absolute_percentage_error: 1807634816.0000 - val_mean_squared_error: 751780.2500\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 202s 2s/step - loss: 939039.6250 - mean_absolute_percentage_error: 2085473408.0000 - mean_squared_error: 939039.6250 - val_loss: 742151.3750 - val_mean_absolute_percentage_error: 2221154304.0000 - val_mean_squared_error: 742151.3750\n",
      "Epoch 4/20\n",
      "125/125 [==============================] - 200s 2s/step - loss: 931105.4375 - mean_absolute_percentage_error: 2403095808.0000 - mean_squared_error: 931105.4375 - val_loss: 735564.8125 - val_mean_absolute_percentage_error: 2246213632.0000 - val_mean_squared_error: 735564.8125\n",
      "Epoch 5/20\n",
      "125/125 [==============================] - 306s 2s/step - loss: 922547.3750 - mean_absolute_percentage_error: 1976390912.0000 - mean_squared_error: 922547.3750 - val_loss: 731285.3125 - val_mean_absolute_percentage_error: 1987039616.0000 - val_mean_squared_error: 731285.3125\n",
      "Epoch 6/20\n",
      "125/125 [==============================] - 298s 2s/step - loss: 913217.3125 - mean_absolute_percentage_error: 1838883584.0000 - mean_squared_error: 913217.3125 - val_loss: 725191.5000 - val_mean_absolute_percentage_error: 1994602880.0000 - val_mean_squared_error: 725191.5000\n",
      "Epoch 7/20\n",
      "125/125 [==============================] - 282s 2s/step - loss: 903062.4375 - mean_absolute_percentage_error: 1825296384.0000 - mean_squared_error: 903062.4375 - val_loss: 719591.6250 - val_mean_absolute_percentage_error: 2078903808.0000 - val_mean_squared_error: 719591.6250\n",
      "Epoch 8/20\n",
      "125/125 [==============================] - 285s 2s/step - loss: 893667.2500 - mean_absolute_percentage_error: 1869494912.0000 - mean_squared_error: 893667.2500 - val_loss: 715490.2500 - val_mean_absolute_percentage_error: 2143448832.0000 - val_mean_squared_error: 715490.2500\n",
      "Epoch 9/20\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "125/125 [==============================] - 215s 2s/step - loss: 886103.7500 - mean_absolute_percentage_error: 1944836224.0000 - mean_squared_error: 886103.7500 - val_loss: 714392.1250 - val_mean_absolute_percentage_error: 2228860416.0000 - val_mean_squared_error: 714392.1250\n",
      "Epoch 10/20\n",
      "125/125 [==============================] - 239s 2s/step - loss: 878934.3750 - mean_absolute_percentage_error: 1969614080.0000 - mean_squared_error: 878934.3750 - val_loss: 724824.2500 - val_mean_absolute_percentage_error: 2008506112.0000 - val_mean_squared_error: 724824.2500\n",
      "Epoch 11/20\n",
      "125/125 [==============================] - 304s 2s/step - loss: 873268.0625 - mean_absolute_percentage_error: 2091314176.0000 - mean_squared_error: 873268.0625 - val_loss: 708983.1250 - val_mean_absolute_percentage_error: 2439692800.0000 - val_mean_squared_error: 708983.1250\n",
      "Epoch 12/20\n",
      "125/125 [==============================] - 384s 3s/step - loss: 866292.3125 - mean_absolute_percentage_error: 2056521600.0000 - mean_squared_error: 866292.3125 - val_loss: 700679.1875 - val_mean_absolute_percentage_error: 2640231168.0000 - val_mean_squared_error: 700679.1875\n",
      "Epoch 13/20\n",
      "125/125 [==============================] - 244s 2s/step - loss: 860071.3750 - mean_absolute_percentage_error: 2072371584.0000 - mean_squared_error: 860071.3750 - val_loss: 697788.7500 - val_mean_absolute_percentage_error: 2579060480.0000 - val_mean_squared_error: 697788.7500\n",
      "Epoch 14/20\n",
      "125/125 [==============================] - 217s 2s/step - loss: 853247.0000 - mean_absolute_percentage_error: 1983100928.0000 - mean_squared_error: 853247.0000 - val_loss: 698752.0625 - val_mean_absolute_percentage_error: 2353897216.0000 - val_mean_squared_error: 698752.0625\n",
      "Epoch 15/20\n",
      "125/125 [==============================] - 237s 2s/step - loss: 848526.9375 - mean_absolute_percentage_error: 1994126336.0000 - mean_squared_error: 848526.9375 - val_loss: 697434.6250 - val_mean_absolute_percentage_error: 2349303296.0000 - val_mean_squared_error: 697434.6250\n",
      "Epoch 16/20\n",
      "125/125 [==============================] - 222s 2s/step - loss: 840840.4375 - mean_absolute_percentage_error: 1862278400.0000 - mean_squared_error: 840840.4375 - val_loss: 694318.3125 - val_mean_absolute_percentage_error: 2195065344.0000 - val_mean_squared_error: 694318.3125\n",
      "Epoch 17/20\n",
      "125/125 [==============================] - 196s 2s/step - loss: 834378.8750 - mean_absolute_percentage_error: 1648996224.0000 - mean_squared_error: 834378.8750 - val_loss: 696554.3125 - val_mean_absolute_percentage_error: 2041305344.0000 - val_mean_squared_error: 696554.3125\n",
      "Epoch 18/20\n",
      "125/125 [==============================] - 194s 2s/step - loss: 828027.2500 - mean_absolute_percentage_error: 1429760000.0000 - mean_squared_error: 828027.2500 - val_loss: 695702.5000 - val_mean_absolute_percentage_error: 2096738944.0000 - val_mean_squared_error: 695702.5000\n",
      "Epoch 1/20\n",
      "125/125 [==============================] - 89s 673ms/step - loss: 995811.0000 - mean_absolute_percentage_error: 164420800.0000 - mean_squared_error: 995811.0000 - val_loss: 802352.1250 - val_mean_absolute_percentage_error: 242388752.0000 - val_mean_squared_error: 802352.1250\n",
      "Epoch 2/20\n",
      "125/125 [==============================] - 84s 670ms/step - loss: 990953.3125 - mean_absolute_percentage_error: 291046592.0000 - mean_squared_error: 990953.3125 - val_loss: 798875.6250 - val_mean_absolute_percentage_error: 331840768.0000 - val_mean_squared_error: 798875.6250\n",
      "Epoch 3/20\n",
      "125/125 [==============================] - 87s 694ms/step - loss: 987965.9375 - mean_absolute_percentage_error: 379780640.0000 - mean_squared_error: 987965.9375 - val_loss: 796027.5625 - val_mean_absolute_percentage_error: 406523552.0000 - val_mean_squared_error: 796027.5625\n",
      "Epoch 4/20\n",
      "125/125 [==============================] - 85s 681ms/step - loss: 985405.6250 - mean_absolute_percentage_error: 453799584.0000 - mean_squared_error: 985405.6250 - val_loss: 793557.7500 - val_mean_absolute_percentage_error: 472368064.0000 - val_mean_squared_error: 793557.7500\n",
      "Epoch 5/20\n",
      "125/125 [==============================] - 92s 734ms/step - loss: 983085.1250 - mean_absolute_percentage_error: 514455648.0000 - mean_squared_error: 983085.1250 - val_loss: 791241.9375 - val_mean_absolute_percentage_error: 535063232.0000 - val_mean_squared_error: 791241.9375\n",
      "Epoch 6/20\n",
      "125/125 [==============================] - 94s 748ms/step - loss: 980917.3125 - mean_absolute_percentage_error: 579061440.0000 - mean_squared_error: 980917.3125 - val_loss: 789080.5625 - val_mean_absolute_percentage_error: 594444288.0000 - val_mean_squared_error: 789080.5625\n",
      "Epoch 7/20\n",
      "125/125 [==============================] - 82s 659ms/step - loss: 978862.8750 - mean_absolute_percentage_error: 638540928.0000 - mean_squared_error: 978862.8750 - val_loss: 787007.6875 - val_mean_absolute_percentage_error: 652224960.0000 - val_mean_squared_error: 787007.6875\n",
      "Epoch 8/20\n",
      "125/125 [==============================] - 84s 669ms/step - loss: 976899.8750 - mean_absolute_percentage_error: 698021568.0000 - mean_squared_error: 976899.8750 - val_loss: 785005.1875 - val_mean_absolute_percentage_error: 708846400.0000 - val_mean_squared_error: 785005.1875\n",
      "Epoch 9/20\n",
      "125/125 [==============================] - 83s 666ms/step - loss: 975008.3750 - mean_absolute_percentage_error: 751721152.0000 - mean_squared_error: 975008.3750 - val_loss: 783108.3750 - val_mean_absolute_percentage_error: 763236416.0000 - val_mean_squared_error: 783108.3750\n",
      "Epoch 10/20\n",
      "125/125 [==============================] - 84s 676ms/step - loss: 973195.0000 - mean_absolute_percentage_error: 813184256.0000 - mean_squared_error: 973195.0000 - val_loss: 781226.3125 - val_mean_absolute_percentage_error: 817964736.0000 - val_mean_squared_error: 781226.3125\n",
      "Epoch 11/20\n",
      "125/125 [==============================] - 82s 660ms/step - loss: 971438.5625 - mean_absolute_percentage_error: 865532416.0000 - mean_squared_error: 971438.5625 - val_loss: 779468.1250 - val_mean_absolute_percentage_error: 869800064.0000 - val_mean_squared_error: 779468.1250\n",
      "Epoch 12/20\n",
      "125/125 [==============================] - 83s 661ms/step - loss: 969740.6250 - mean_absolute_percentage_error: 918181184.0000 - mean_squared_error: 969740.6250 - val_loss: 777715.7500 - val_mean_absolute_percentage_error: 922170496.0000 - val_mean_squared_error: 777715.7500\n",
      "Epoch 13/20\n",
      "125/125 [==============================] - 87s 697ms/step - loss: 968085.9375 - mean_absolute_percentage_error: 973073024.0000 - mean_squared_error: 968085.9375 - val_loss: 776040.0000 - val_mean_absolute_percentage_error: 972922816.0000 - val_mean_squared_error: 776040.0000\n",
      "Epoch 14/20\n",
      "125/125 [==============================] - 103s 826ms/step - loss: 966476.8750 - mean_absolute_percentage_error: 1024080384.0000 - mean_squared_error: 966476.8750 - val_loss: 774384.9375 - val_mean_absolute_percentage_error: 1023737216.0000 - val_mean_squared_error: 774384.9375\n",
      "Epoch 15/20\n",
      "125/125 [==============================] - 102s 821ms/step - loss: 964904.7500 - mean_absolute_percentage_error: 1073153280.0000 - mean_squared_error: 964904.7500 - val_loss: 772757.6250 - val_mean_absolute_percentage_error: 1074403072.0000 - val_mean_squared_error: 772757.6250\n",
      "Epoch 16/20\n",
      "125/125 [==============================] - 130s 1s/step - loss: 963366.2500 - mean_absolute_percentage_error: 1124568704.0000 - mean_squared_error: 963366.2500 - val_loss: 771207.6875 - val_mean_absolute_percentage_error: 1123351040.0000 - val_mean_squared_error: 771207.6875\n",
      "Epoch 17/20\n",
      "125/125 [==============================] - 98s 785ms/step - loss: 961873.9375 - mean_absolute_percentage_error: 1179699968.0000 - mean_squared_error: 961873.9375 - val_loss: 769654.4375 - val_mean_absolute_percentage_error: 1173076480.0000 - val_mean_squared_error: 769654.4375\n",
      "Epoch 18/20\n",
      "125/125 [==============================] - 99s 796ms/step - loss: 960359.7500 - mean_absolute_percentage_error: 1226586624.0000 - mean_squared_error: 960359.7500 - val_loss: 768007.8750 - val_mean_absolute_percentage_error: 1225350016.0000 - val_mean_squared_error: 768007.8750\n",
      "Epoch 19/20\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "125/125 [==============================] - 104s 829ms/step - loss: 958477.9375 - mean_absolute_percentage_error: 1096582784.0000 - mean_squared_error: 958477.9375 - val_loss: 766140.8750 - val_mean_absolute_percentage_error: 1074234624.0000 - val_mean_squared_error: 766140.8750\n",
      "Epoch 20/20\n",
      "125/125 [==============================] - 119s 955ms/step - loss: 956143.2500 - mean_absolute_percentage_error: 601336704.0000 - mean_squared_error: 956143.2500 - val_loss: 764702.8750 - val_mean_absolute_percentage_error: 921368384.0000 - val_mean_squared_error: 764702.8750\n"
     ]
    }
   ],
   "source": [
    "#########\n",
    "epochs = 20\n",
    "batch_size = 64\n",
    " \n",
    "history1a = model1a.fit(X, Y2, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n",
    "history2a = model2a.fit(X, Y2, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n",
    "history3a = model3a.fit(X, Y2, epochs=epochs, batch_size=batch_size,validation_split=0.2, shuffle=True,\n",
    "                    callbacks=[tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=2, min_delta=0.000001)])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "c2db4c59",
   "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.title('Mean squared error')\n",
    "plt.xlabel('epoch')\n",
    "plt.plot(history1a.history['mean_squared_error'], label='train_lstm', color = \"red\")\n",
    "plt.plot(history1a.history['val_mean_squared_error'], label='test_lstm', linestyle='--', color = \"red\")\n",
    "plt.plot(history2a.history['mean_squared_error'], label='train_bidirect', color = \"navy\")\n",
    "plt.plot(history2a.history['val_mean_squared_error'], label='test_bidirect', linestyle='--', color = \"navy\")\n",
    "plt.plot(history3a.history['mean_squared_error'], label='train_stack', color = \"gold\")\n",
    "plt.plot(history3a.history['val_mean_squared_error'], label='test_stack', linestyle='--', color = \"gold\")\n",
    "plt.legend(bbox_to_anchor = (1.05, 0.6))\n",
    "plt.show()\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "da252f60",
   "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.title('Mean absolute percentage error (MAPE)')\n",
    "plt.xlabel('epoch')\n",
    "plt.plot(history1a.history['mean_absolute_percentage_error'], label='train_lstm', color = \"red\")\n",
    "plt.plot(history1a.history['val_mean_absolute_percentage_error'], label='test_lstm', linestyle='--', color = \"red\")\n",
    "plt.plot(history2a.history['mean_absolute_percentage_error'], label='train_bidirect', color = \"navy\")\n",
    "plt.plot(history2a.history['val_mean_absolute_percentage_error'], label='test_bidirect', linestyle='--', color = \"navy\")\n",
    "plt.plot(history3a.history['mean_absolute_percentage_error'], label='train_stack', color = \"gold\")\n",
    "plt.plot(history3a.history['val_mean_absolute_percentage_error'], label='test_stack', linestyle='--', color = \"gold\")\n",
    "plt.legend(bbox_to_anchor = (1.05, 0.6))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "9352d351",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "828027.25"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history2a.history['mean_squared_error'][-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "id": "f210e4ec",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR MSE: 1010437.907273\n"
     ]
    }
   ],
   "source": [
    "#Train a linear regression model with the non-text features to compare with the results of LSTM models\n",
    "\n",
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn import metrics\n",
    "\n",
    "X0 = df[[\"clicksCount\", \"orderNum\", \"originalFlag\"]]\n",
    "Y2 = df['likeCount'].values\n",
    "\n",
    "\n",
    "\n",
    "X0_train, X0_test, Y2_train, Y2_test = train_test_split(X,Y2, test_size = 0.20, random_state = 42)\n",
    "\n",
    "\n",
    "lr = LinearRegression()\n",
    "lr.fit(X0_train, Y2_train)\n",
    "y_pred = lr.predict(X0_test)\n",
    "\n",
    "mse_lr=metrics.mean_squared_error(Y2_test, y_pred)\n",
    "mape_lr = metrics.mean_absolute_percentage_error(Y2_test, y_pred)\n",
    "\n",
    "print(\"LR MSE: %f\" % (mse_lr))\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "id": "0a0bd4ad",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SVM accuracy: 0.669835\n"
     ]
    }
   ],
   "source": [
    "from sklearn.svm import SVC\n",
    "\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "\n",
    "X0 = df[[\"clicksCount\", \"orderNum\", \"originalFlag\"]]\n",
    "Y2 = df['advertisement']\n",
    "\n",
    "\n",
    "X0_train, X0_test, Y2_train, Y2_test = train_test_split(X0,Y2, test_size = 0.20, random_state = 42)\n",
    "\n",
    "scaling = MinMaxScaler(feature_range=(-1,1)).fit(X0_train)\n",
    "X0_train = scaling.transform(X0_train)\n",
    "X0_test = scaling.transform(X0_test)\n",
    "\n",
    "regressor= SVC(kernel='linear')\n",
    "regressor.fit(X0_train,Y2_train)\n",
    "y_pred_svm=regressor.predict(X0_test)\n",
    "#y_pred_svm = cross_val_predict(regressor, x, y)\n",
    "\n",
    "svm_mse=metrics.accuracy_score(Y2_test, y_pred_svm)\n",
    "print(\"SVM accuracy: %f\" % (svm_mse))\n",
    "\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}