Recurrent Neural Network.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "a383dbf9",
   "metadata": {},
   "source": [
    "# Part 5: Recurrent Neural Network\n",
    "\n",
    "### In this section, the cleaned text will be converted to a suitable form using GloVe embedding, then used to train a RNN with LSTM layer.\n",
    "\n",
    "### This section will also explore and test different methods of fine-tuning the RNN and evaluating its effectiveness"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b26f6696",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "!{sys.executable} -m pip install pandas_profiling\n",
    "!{sys.executable} -m pip install spacy\n",
    "!{sys.executable} -m pip install nltk\n",
    "!{sys.executable} -m pip install imblearn\n",
    "!{sys.executable} -m pip install scikit-learn-intelex\n",
    "!{sys.executable} -m pip install tensorflow==2.9.0 --user\n",
    "!{sys.executable} -m spacy download en_core_web_md\n",
    "!{sys.executable} -m pip install wget\n",
    "!{sys.executable} -m pip install unzip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "880c273d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Saved under glove.6B.zip\n"
     ]
    }
   ],
   "source": [
    "!{sys.executable} -m wget http://nlp.stanford.edu/data/glove.6B.zip"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2a5cad6f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.9.0\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sb\n",
    "import spacy\n",
    "import nltk\n",
    "import re\n",
    "from sklearn import metrics\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "from imblearn.over_sampling import ADASYN, SMOTE\n",
    "from sklearn.feature_selection import SelectFromModel\n",
    "from sklearn.feature_selection import VarianceThreshold\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import tensorflow as tf\n",
    "from itertools import product\n",
    "from tensorflow.keras.layers import Embedding\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.preprocessing.text import one_hot\n",
    "from tensorflow.keras.layers import LSTM\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.layers import Bidirectional\n",
    "from tensorflow.keras.layers import Dropout,Conv1D, MaxPooling1D, GlobalMaxPooling1D, LeakyReLU\n",
    "from keras.regularizers import l2\n",
    "from tensorflow.keras.preprocessing.text import one_hot, Tokenizer\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from keras.preprocessing.text import Tokenizer\n",
    "\n",
    "adasyn=ADASYN()\n",
    "sb.set()\n",
    "\n",
    "\"\"\" NOTE: SET YOUR PROJECT ROOT DIRECTORY HERE \"\"\"\n",
    "PROJ_DIR = r\"\"\n",
    "RANDOM_SPLIT_SEED = 11\n",
    "MAX_FEATURES = 50000\n",
    "MAX_LEN = 200\n",
    "EMBEDDING_DIM = 100\n",
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc52ad5c",
   "metadata": {},
   "source": [
    "Use CUDA DNN acceleration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "115410cd",
   "metadata": {},
   "outputs": [],
   "source": [
    "physical_devices = tf.config.list_physical_devices('GPU')\n",
    "tf.config.experimental.set_memory_growth(physical_devices[0], enable=True)\n",
    "config = tf.compat.v1.ConfigProto()\n",
    "config.gpu_options.allow_growth = True\n",
    "sess = tf.compat.v1.Session(config=config)\n",
    "tf.compat.v1.keras.backend.set_session(sess)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c3cc4265",
   "metadata": {},
   "source": [
    "#### The pre-processed text is read from csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f9461b36",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(PROJ_DIR + r\"\\sc1015-project\\dataset\\fake_job_postings.csv\")\n",
    "\n",
    "cleaned_text = pd.read_csv('lemmatized_text.csv').squeeze() # convert to pd series\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d074d3aa",
   "metadata": {},
   "source": [
    "#### Use GloVe model for embedding layer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b85cb4f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "embeddings_index = {}\n",
    "with open('glove_embeddings/glove.6B.100d.txt', 'r', encoding='utf-8') as f:\n",
    "    for line in f:\n",
    "        values = line.split()\n",
    "        word = values[0]\n",
    "        vector = np.asarray(values[1:], dtype='float32')\n",
    "        embeddings_index[word] = vector"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e9c6d556",
   "metadata": {},
   "source": [
    "#### Next, the text will be tokenized"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "e115276c",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = Tokenizer(num_words=MAX_FEATURES)\n",
    "tokenizer.fit_on_texts(cleaned_text)\n",
    "word_index = tokenizer.word_index\n",
    "sequences = tokenizer.texts_to_sequences(cleaned_text)\n",
    "ml_seq = pad_sequences(sequences, maxlen=MAX_LEN)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0f94f03b",
   "metadata": {},
   "source": [
    "#### Form embedding matrix to be used in the RNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "1312f770",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_words = min(MAX_FEATURES, len(word_index) + 1)\n",
    "embedding_matrix = np.zeros((num_words, EMBEDDING_DIM))\n",
    "for word, i in word_index.items():\n",
    "    if i >= MAX_FEATURES:\n",
    "        continue\n",
    "    embedding_vector = embeddings_index.get(word)\n",
    "    if embedding_vector is not None:\n",
    "        embedding_matrix[i] = embedding_vector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7b051bc9",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "0507f0ab",
   "metadata": {},
   "source": [
    "#### The resulting padded data is split and the train set is resampled with ADASYN to compensate class imbalance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "1572007e",
   "metadata": {},
   "outputs": [],
   "source": [
    "sss=StratifiedShuffleSplit(n_splits=1, random_state=RANDOM_SPLIT_SEED, test_size=0.2)\n",
    "X=np.zeros(shape=ml_seq.shape[0],dtype=np.bool_)\n",
    "y=np.array(df['fraudulent']) # prediction target\n",
    "for train_index, test_index in sss.split(X, y):\n",
    "    X_train, X_test = ml_seq[train_index,:], ml_seq[test_index,:]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "\n",
    "X_res, y_res = adasyn.fit_resample(X_train, y_train)\n",
    "#smote = SMOTE(random_state=42)\n",
    "#X_resampled, y_resampled = smote.fit_resample(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "412482e1",
   "metadata": {},
   "source": [
    "### This section defines the RNN and its output layers, as well as tests the effectiveness of different number of LSTM units.\n",
    "**The embedded layer uses a non-trainable matrix as defined previously, derived from the GloVe model,\n",
    "followed by a bi-directional LSTM to help with memory in time domain,\n",
    "followed by a max pooling layer to summarize max vals in time domain,\n",
    "then 3 successive layers of dense and dropout\n",
    "The output activation is sigmoid so it can be rounded to a boolean value.\n",
    "LeakyRelu activation function was used to mitigate the dying Relu problem**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5ffc75c8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      " Layer (type)                Output Shape              Param #   \n",
      "=================================================================\n",
      " embedding_3 (Embedding)     (None, 200, 100)          5000000   \n",
      "                                                                 \n",
      " bidirectional_3 (Bidirectio  (None, 200, 256)         234496    \n",
      " nal)                                                            \n",
      "                                                                 \n",
      " global_max_pooling1d_3 (Glo  (None, 256)              0         \n",
      " balMaxPooling1D)                                                \n",
      "                                                                 \n",
      " dense_12 (Dense)            (None, 256)               65792     \n",
      "                                                                 \n",
      " leaky_re_lu_9 (LeakyReLU)   (None, 256)               0         \n",
      "                                                                 \n",
      " dropout_9 (Dropout)         (None, 256)               0         \n",
      "                                                                 \n",
      " dense_13 (Dense)            (None, 128)               32896     \n",
      "                                                                 \n",
      " leaky_re_lu_10 (LeakyReLU)  (None, 128)               0         \n",
      "                                                                 \n",
      " dropout_10 (Dropout)        (None, 128)               0         \n",
      "                                                                 \n",
      " dense_14 (Dense)            (None, 64)                8256      \n",
      "                                                                 \n",
      " leaky_re_lu_11 (LeakyReLU)  (None, 64)                0         \n",
      "                                                                 \n",
      " dropout_11 (Dropout)        (None, 64)                0         \n",
      "                                                                 \n",
      " dense_15 (Dense)            (None, 1)                 65        \n",
      "                                                                 \n",
      "=================================================================\n",
      "Total params: 5,341,505\n",
      "Trainable params: 341,505\n",
      "Non-trainable params: 5,000,000\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(MAX_FEATURES, EMBEDDING_DIM, weights=[embedding_matrix], input_length=MAX_LEN, trainable=False))\n",
    "model.add(Bidirectional(LSTM(128,return_sequences=True)))\n",
    "model.add(GlobalMaxPooling1D())\n",
    "model.add(Dense(256))\n",
    "model.add(LeakyReLU(alpha=0.05))\n",
    "model.add(Dropout(0.3))\n",
    "model.add(Dense(128))\n",
    "model.add(LeakyReLU(alpha=0.05))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(64))\n",
    "model.add(LeakyReLU(alpha=0.05))\n",
    "model.add(Dropout(0.1))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.compile(optimizer='adam',\n",
    "              loss='binary_crossentropy',\n",
    "              metrics=['accuracy'])\n",
    "print(model.summary())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "d0d9e420",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.1409 - accuracy: 0.9444\n",
      "Epoch 1: val_loss improved from inf to 0.17023, saving model to rnn_weights.h5\n",
      "423/423 [==============================] - 59s 126ms/step - loss: 0.1409 - accuracy: 0.9444 - val_loss: 0.1702 - val_accuracy: 0.9360\n",
      "Epoch 2/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0576 - accuracy: 0.9822\n",
      "Epoch 2: val_loss improved from 0.17023 to 0.07313, saving model to rnn_weights.h5\n",
      "423/423 [==============================] - 62s 145ms/step - loss: 0.0576 - accuracy: 0.9822 - val_loss: 0.0731 - val_accuracy: 0.9787\n",
      "Epoch 3/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0388 - accuracy: 0.9881\n",
      "Epoch 3: val_loss did not improve from 0.07313\n",
      "423/423 [==============================] - 55s 129ms/step - loss: 0.0388 - accuracy: 0.9881 - val_loss: 0.1052 - val_accuracy: 0.9737\n",
      "Epoch 4/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0310 - accuracy: 0.9905\n",
      "Epoch 4: val_loss did not improve from 0.07313\n",
      "423/423 [==============================] - 58s 137ms/step - loss: 0.0310 - accuracy: 0.9905 - val_loss: 0.0743 - val_accuracy: 0.9824\n",
      "Epoch 5/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0234 - accuracy: 0.9922\n",
      "Epoch 5: val_loss did not improve from 0.07313\n",
      "423/423 [==============================] - 54s 127ms/step - loss: 0.0234 - accuracy: 0.9922 - val_loss: 0.0750 - val_accuracy: 0.9818\n",
      "Epoch 6/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0159 - accuracy: 0.9942\n",
      "Epoch 6: val_loss improved from 0.07313 to 0.05939, saving model to rnn_weights.h5\n",
      "423/423 [==============================] - 53s 126ms/step - loss: 0.0159 - accuracy: 0.9942 - val_loss: 0.0594 - val_accuracy: 0.9835\n",
      "Epoch 7/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0105 - accuracy: 0.9966\n",
      "Epoch 7: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 54s 127ms/step - loss: 0.0105 - accuracy: 0.9966 - val_loss: 0.0652 - val_accuracy: 0.9827\n",
      "Epoch 8/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0105 - accuracy: 0.9967\n",
      "Epoch 8: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 56s 133ms/step - loss: 0.0105 - accuracy: 0.9967 - val_loss: 0.0764 - val_accuracy: 0.9832\n",
      "Epoch 9/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0101 - accuracy: 0.9964\n",
      "Epoch 9: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 58s 136ms/step - loss: 0.0101 - accuracy: 0.9964 - val_loss: 0.0830 - val_accuracy: 0.9821\n",
      "Epoch 10/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0081 - accuracy: 0.9978\n",
      "Epoch 10: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 55s 129ms/step - loss: 0.0081 - accuracy: 0.9978 - val_loss: 0.0746 - val_accuracy: 0.9821\n",
      "Epoch 11/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0085 - accuracy: 0.9968\n",
      "Epoch 11: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 58s 137ms/step - loss: 0.0085 - accuracy: 0.9968 - val_loss: 0.0684 - val_accuracy: 0.9841\n",
      "Epoch 12/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0056 - accuracy: 0.9981\n",
      "Epoch 12: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 56s 131ms/step - loss: 0.0056 - accuracy: 0.9981 - val_loss: 0.0908 - val_accuracy: 0.9832\n",
      "Epoch 13/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0039 - accuracy: 0.9986\n",
      "Epoch 13: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 54s 127ms/step - loss: 0.0039 - accuracy: 0.9986 - val_loss: 0.0869 - val_accuracy: 0.9793\n",
      "Epoch 14/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0055 - accuracy: 0.9982\n",
      "Epoch 14: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 55s 131ms/step - loss: 0.0055 - accuracy: 0.9982 - val_loss: 0.1022 - val_accuracy: 0.9813\n",
      "Epoch 15/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0057 - accuracy: 0.9979\n",
      "Epoch 15: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 56s 132ms/step - loss: 0.0057 - accuracy: 0.9979 - val_loss: 0.1161 - val_accuracy: 0.9827\n",
      "Epoch 16/50\n",
      "423/423 [==============================] - ETA: 0s - loss: 0.0060 - accuracy: 0.9981\n",
      "Epoch 16: val_loss did not improve from 0.05939\n",
      "423/423 [==============================] - 57s 135ms/step - loss: 0.0060 - accuracy: 0.9981 - val_loss: 0.0948 - val_accuracy: 0.9860\n",
      "Epoch 16: early stopping\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Define the callbacks\n",
    "early_stop = EarlyStopping(monitor='val_loss', patience=10, verbose=1, mode='min')\n",
    "model_checkpoint = ModelCheckpoint('rnn_weights.h5', save_best_only=True, save_weights_only=True, monitor='val_loss', mode='min', verbose=1)\n",
    "\n",
    "hist = model.fit(X_res,y_res,validation_data=(X_test,y_test),epochs=50,batch_size=64, callbacks=[early_stop, model_checkpoint])\n",
    "# Load the best model weights\n",
    "model.load_weights('rnn_weights.h5')\n",
    "\n",
    "plt.plot(hist.history['val_loss'], color='b', label=\"validation loss\")\n",
    "plt.plot(hist.history['loss'], color='red', label=\"loss\")\n",
    "plt.title(\"Model Loss\")\n",
    "plt.xlabel(\"Number of Epochs\")\n",
    "plt.ylabel(\"Loss\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "2ca2eb50",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "56/56 [==============================] - 4s 46ms/step\n",
      "RNN test accuracy score : 0.9835011185682326\n"
     ]
    }
   ],
   "source": [
    "test_probs = model.predict(X_test, batch_size=64)\n",
    "test_preds = (test_probs >= 0.5).astype(int)\n",
    "\n",
    "test_accuracy = accuracy_score(y_test, test_preds)\n",
    "\n",
    "print(f\"RNN test accuracy score : {test_accuracy}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "d4facb66",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True Positive Rate: 0.7687861271676301\n",
      "True Negative Rate: 0.9944166911548633\n",
      "False Positive Rate: 0.005583308845136644\n",
      "False Negative Rate: 0.23121387283236994\n",
      "Accuracy: 0.9835011185682326\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.99      0.99      0.99      3403\n",
      "           1       0.88      0.77      0.82       173\n",
      "\n",
      "    accuracy                           0.98      3576\n",
      "   macro avg       0.93      0.88      0.90      3576\n",
      "weighted avg       0.98      0.98      0.98      3576\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Calculate the confusion matrix\n",
    "cm = confusion_matrix(y_test, test_preds)\n",
    "\n",
    "# Define the class labels\n",
    "class_names = ['Real', 'Fake']\n",
    "\n",
    "# Create a heatmap of the confusion matrix\n",
    "sb.heatmap(cm, annot=True, fmt='d', cmap='Blues', xticklabels=class_names, yticklabels=class_names)\n",
    "\n",
    "# Set the axis labels and title\n",
    "plt.xlabel('Predicted')\n",
    "plt.ylabel('Actual')\n",
    "plt.title('Confusion Matrix')\n",
    "\n",
    "# Display the plot\n",
    "plt.show()\n",
    "\n",
    "TP = cm[1, 1]\n",
    "TN = cm[0, 0]\n",
    "FP = cm[0, 1]\n",
    "FN = cm[1, 0]\n",
    "\n",
    "TPR = TP / (TP + FN)\n",
    "print(\"True Positive Rate:\", TPR)\n",
    "TNR = TN / (TN + FP) \n",
    "print(\"True Negative Rate:\", TNR)\n",
    "FPR = FP / (FP + TN)\n",
    "print(\"False Positive Rate:\", FPR)\n",
    "FNR = FN / (TP + FN)\n",
    "print(\"False Negative Rate:\", FNR)\n",
    "ACC = (TP + TN) / (TP + FP + FN + TN)\n",
    "print(\"Accuracy:\", ACC)\n",
    "\n",
    "print(metrics.classification_report(y_test, test_preds))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ffc0ac9",
   "metadata": {},
   "source": [
    "### Conclusion for Recurrent Neural Network model\n",
    "\n",
    "**In conclusion, the best overall performance was achieved using a single LSTM layer of 128 units, yielding f1-score of 0.99/0.82, 0.99 TNR, 0.77 TPR. Addition of the globalmaxpooling and dropout/dense layers helped with improving the generalization of the model.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fe5f7edc",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}