train_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train Model\n",
    "\n",
    "This notebook defines the model, trains it, and then evaluates it on the validation and test sets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Import libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM\n",
    "from keras.layers import Dense\n",
    "from keras.regularizers import l2\n",
    "from keras.layers import TimeDistributed, Bidirectional, BatchNormalization, Dropout, Input, Add, Masking\n",
    "from keras import Model\n",
    "import pdb\n",
    "import pandas as pd\n",
    "from keras.callbacks import EarlyStopping\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from keras.models import load_model\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "import glob\n",
    "import sys\n",
    "from sklearn.metrics import auc\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define some constants related to the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_SIZE = 64\n",
    "INPUT_SEQ_LEN_MODEL1 = 10\n",
    "INPUT_NUM_CH_MODEL1 = 5\n",
    "INPUT_FEATS_MODEL2 = 33"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define a function to calculate the ROC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def roc(predictions, true, filename):\n",
    "    predictions = predictions.flatten()\n",
    "    true = true.flatten()\n",
    "\n",
    "    thresh_vals = np.linspace(np.min(predictions), np.max(predictions), 50)\n",
    "    results = []\n",
    "    for thresh in thresh_vals:\n",
    "        tmp_predictions = (predictions > thresh).astype(int)\n",
    "        tn, fp, fn, tp = confusion_matrix(true, tmp_predictions).ravel()\n",
    "        tpr = tp/(tp+fn)\n",
    "        fpr = fp/(tn+fp)\n",
    "        acc = (tp+tn)/(tn+fp+fn+tp)\n",
    "\n",
    "        tmp_dict = {'acc': acc, 'tpr': tpr, 'fpr': fpr, 'thresh': thresh}\n",
    "        results.append(tmp_dict)\n",
    "\n",
    "    results = pd.DataFrame(results)\n",
    "    results = results.sort_values(by='thresh', ascending=False)\n",
    "    \n",
    "    #calculate the AUC\n",
    "    AUC = auc(results['fpr'].values, results['tpr'].values)\n",
    "\n",
    "    plt.figure(figsize=(10,6))\n",
    "    plt.plot(results['fpr'], results['tpr'], '*-')\n",
    "    plt.xlabel('FPR')\n",
    "    plt.ylabel('TPR')\n",
    "    plt.title('ROC\\nAUC=%.2f' % AUC)\n",
    "    plt.tight_layout()\n",
    "    plt.savefig('plots/ROC_%s.png' % filename, dpi=250)\n",
    "    plt.show()\n",
    "\n",
    "    results = results.sort_values(by='acc', ascending=False)\n",
    "    final_thresh = results.head(1)['thresh'].values[0]\n",
    "\n",
    "    return results, final_thresh\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the data and set up the train, validation, and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#load the training data\n",
    "train = pd.read_pickle('train.pkl').reset_index(drop=True)\n",
    "\n",
    "#define the validation set as 15% of the training set\n",
    "N = len(train)\n",
    "train = train.sample(frac=1).reset_index(drop=True)\n",
    "val = train[:int(N*0.2)]\n",
    "train = train[int(N*0.2):]\n",
    "\n",
    "#separate out the training data\n",
    "#note that input data for LSTM should be in the format of (number_samples, sequence_len, num_channels)\n",
    "y_train = np.asarray(list(train['label']))\n",
    "y_train = to_categorical(y_train)\n",
    "X_train_cont = np.asarray(list(train['X_cont'].values))\n",
    "X_train_cat = train.drop(['X_cont', 'label'], axis=1)\n",
    "\n",
    "#for the mask layer, any NaN values need to be replaced by a unique value. use the constant np.pi\n",
    "X_train_cat[X_train_cat.isna()] = np.pi\n",
    "X_train_cat = np.asarray(list(X_train_cat.values))\n",
    "\n",
    "#separate out the training data\n",
    "y_val = np.asarray(list(val['label']))\n",
    "y_val = to_categorical(y_val)\n",
    "X_val_cont = np.asarray(list(val['X_cont'].values))\n",
    "X_val_cat = val.drop(['X_cont', 'label'], axis=1)\n",
    "\n",
    "#replace NaN with pi\n",
    "X_val_cat[X_val_cat.isna()] = np.pi\n",
    "X_val_cat = np.asarray(list(X_val_cat.values))\n",
    "\n",
    "#load the test data\n",
    "test = pd.read_pickle('test.pkl')\n",
    "y_test = np.asarray(list(test['label']))\n",
    "y_test = to_categorical(y_test)\n",
    "X_test_cont = np.asarray(list(test['X_cont'].values))\n",
    "X_test_cat = test.drop(['X_cont', 'label'], axis=1)\n",
    "\n",
    "#replace NaN with pi\n",
    "X_test_cat[X_test_cat.isna()] = np.pi\n",
    "X_test_cat = np.asarray(list(X_test_cat.values))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Calculate the class weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "class weights -- no sepsis: 1.0, sepsis: 54.361174458380844\n"
     ]
    }
   ],
   "source": [
    "count_class_0 = np.sum(y_train[:,1].astype(int)==0)\n",
    "count_class_1 = np.sum(y_train[:, 1].astype(int) == 1)\n",
    "max_class_counts = np.max((count_class_0, count_class_1))\n",
    "class_weights = {0: max_class_counts/count_class_0, 1: max_class_counts/count_class_1}\n",
    "print('class weights -- no sepsis: {}, sepsis: {}'.format(class_weights[0], class_weights[1]))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the model using the functional API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_1\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 10, 5)        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional_1 (Bidirectional) (None, 10, 200)      84800       input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "input_2 (InputLayer)            (None, 33)           0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional_2 (Bidirectional) (None, 150)          165600      bidirectional_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "masking_1 (Masking)             (None, 33)           0           input_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 35)           5285        bidirectional_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 30)           1020        masking_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 35)           140         dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 30)           120         dense_3[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 15)           540         batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 15)           465         batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 15)           60          dense_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 15)           60          dense_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "add_1 (Add)                     (None, 15)           0           batch_normalization_2[0][0]      \n",
      "                                                                 batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 2)            32          add_1[0][0]                      \n",
      "==================================================================================================\n",
      "Total params: 258,122\n",
      "Trainable params: 257,932\n",
      "Non-trainable params: 190\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "input1 = Input(shape=(INPUT_SEQ_LEN_MODEL1, INPUT_NUM_CH_MODEL1))\n",
    "model1 = Bidirectional(LSTM(100, kernel_regularizer=l2(0.001), return_sequences=True))(input1)\n",
    "model1 = Bidirectional(LSTM(75, kernel_regularizer=l2(0.001)))(model1)\n",
    "model1 = Dense(35, kernel_regularizer=l2(0.001), activation='relu')(model1)\n",
    "model1 = BatchNormalization()(model1)\n",
    "model1 = Dense(15, kernel_regularizer=l2(0.001), activation='relu')(model1)\n",
    "model1 = BatchNormalization()(model1)\n",
    "\n",
    "input2 = Input(shape=(INPUT_FEATS_MODEL2,))\n",
    "model2 = Masking(mask_value=np.pi)(input2)\n",
    "model2 = Dense(30, kernel_regularizer=l2(0.001), activation='relu')(model2)\n",
    "model2 = BatchNormalization()(model2)\n",
    "model2 = Dense(15, kernel_regularizer=l2(0.001), activation='relu')(model2)\n",
    "model2 = BatchNormalization()(model2)\n",
    "\n",
    "model_add = Add()([model1, model2])\n",
    "output = Dense(2, kernel_regularizer=l2(0.001), activation='softmax')(model_add)\n",
    "\n",
    "model = Model(inputs=[input1, input2], outputs=output)\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam')\n",
    "print(model.summary())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 776828 samples, validate on 194206 samples\n",
      "Epoch 1/50\n",
      "776828/776828 [==============================] - 292s 375us/step - loss: 1.3216 - val_loss: 0.6175\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 0.61751, saving model to model.h5\n",
      "Epoch 2/50\n",
      "776828/776828 [==============================] - 292s 376us/step - loss: 1.1951 - val_loss: 0.1373\n",
      "\n",
      "Epoch 00002: val_loss improved from 0.61751 to 0.13727, saving model to model.h5\n",
      "Epoch 3/50\n",
      "776828/776828 [==============================] - 301s 388us/step - loss: 1.1605 - val_loss: 1.0921\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 0.13727\n",
      "Epoch 4/50\n",
      "776828/776828 [==============================] - 299s 384us/step - loss: 1.1537 - val_loss: 0.1961\n",
      "\n",
      "Epoch 00004: val_loss did not improve from 0.13727\n",
      "Epoch 5/50\n",
      "776828/776828 [==============================] - 291s 375us/step - loss: 1.1670 - val_loss: 0.6232\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 0.13727\n",
      "Epoch 6/50\n",
      "776828/776828 [==============================] - 285s 367us/step - loss: 1.1526 - val_loss: 0.6367\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 0.13727\n",
      "Epoch 7/50\n",
      "776828/776828 [==============================] - 290s 373us/step - loss: 1.1507 - val_loss: 0.6730\n",
      "\n",
      "Epoch 00007: val_loss did not improve from 0.13727\n"
     ]
    }
   ],
   "source": [
    "checkpoint = ModelCheckpoint('model.h5', monitor='val_loss', verbose=1, save_best_only=True, mode='auto', period=1)\n",
    "earlystop = EarlyStopping(monitor='val_loss', min_delta=0, patience=5)\n",
    "history = model.fit([X_train_cont, X_train_cat],\n",
    "                    y_train,\n",
    "                    batch_size=BATCH_SIZE,\n",
    "                    epochs=50,\n",
    "                    validation_data=([X_val_cont, X_val_cat], y_val),\n",
    "                    callbacks=[earlystop, checkpoint],\n",
    "                    class_weight=class_weights,\n",
    "                    verbose=1)\n",
    "\n",
    "#save the history\n",
    "pickle.dump(history, open('history.pkl', 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Plot the training/validation loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation loss values\n",
    "plt.figure(figsize=(10,6))\n",
    "plt.plot(history.history['loss'], '*-')\n",
    "plt.plot(history.history['val_loss'], '*-')\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Validation'], loc='upper left')\n",
    "plt.tight_layout()\n",
    "plt.savefig('plots/loss.png', dpi=250)\n",
    "plt.show()\n",
    "plt.close()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluate the model on the validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = model.predict([X_val_cont, X_val_cat])\n",
    "\n",
    "results_df, thresh_final = roc(predictions[:,1].flatten(), y_val[:,1].flatten(), 'val')\n",
    "results_df = results_df.sort_values(by='fpr')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Evaluate the model on the test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "predictions = model.predict([X_test_cont, X_test_cat])\n",
    "\n",
    "results_df, thresh_final = roc(predictions[:,1].flatten(), y_test[:,1].flatten(), 'test')\n",
    "results_df = results_df.sort_values(by='fpr')\n"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}