RUL Classification -3 - FINAL.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "388035b4",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-02-09 23:41:50.623411: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_1 (InputLayer)           [(None, 50, 16)]     0           []                               \n",
      "                                                                                                  \n",
      " conv1d (Conv1D)                (None, 45, 50)       4850        ['input_1[0][0]']                \n",
      "                                                                                                  \n",
      " conv1d_1 (Conv1D)              (None, 40, 50)       15050       ['conv1d[0][0]']                 \n",
      "                                                                                                  \n",
      " conv1d_2 (Conv1D)              (None, 40, 50)       15050       ['conv1d[0][0]']                 \n",
      "                                                                                                  \n",
      " add (Add)                      (None, 40, 50)       0           ['conv1d_1[0][0]',               \n",
      "                                                                  'conv1d_2[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling1d (MaxPooling1D)   (None, 13, 50)       0           ['add[0][0]']                    \n",
      "                                                                                                  \n",
      " conv1d_3 (Conv1D)              (None, 8, 100)       30100       ['max_pooling1d[0][0]']          \n",
      "                                                                                                  \n",
      " conv1d_4 (Conv1D)              (None, 3, 100)       60100       ['conv1d_3[0][0]']               \n",
      "                                                                                                  \n",
      " max_pooling1d_1 (MaxPooling1D)  (None, 1, 100)      0           ['conv1d_4[0][0]']               \n",
      "                                                                                                  \n",
      " dropout (Dropout)              (None, 1, 100)       0           ['max_pooling1d_1[0][0]']        \n",
      "                                                                                                  \n",
      " lstm (LSTM)                    (None, 1, 50)        30200       ['dropout[0][0]']                \n",
      "                                                                                                  \n",
      " lstm_1 (LSTM)                  (None, 1, 100)       60400       ['lstm[0][0]']                   \n",
      "                                                                                                  \n",
      " lstm_3 (LSTM)                  (None, 1, 100)       80400       ['lstm_1[0][0]']                 \n",
      "                                                                                                  \n",
      " global_average_pooling1d (Glob  (None, 100)         0           ['lstm_3[0][0]']                 \n",
      " alAveragePooling1D)                                                                              \n",
      "                                                                                                  \n",
      " dropout_1 (Dropout)            (None, 100)          0           ['global_average_pooling1d[0][0]'\n",
      "                                                                 ]                                \n",
      "                                                                                                  \n",
      " dense (Dense)                  (None, 1)            101         ['dropout_1[0][0]']              \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 296,251\n",
      "Trainable params: 296,251\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "import keras\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import os\n",
    "import seaborn as sns\n",
    "# Setting seed for reproducibility\n",
    "np.random.seed(1234)  \n",
    "PYTHONHASHSEED = 0\n",
    "\n",
    "from sklearn import preprocessing\n",
    "from sklearn.metrics import confusion_matrix, recall_score, precision_score\n",
    "from keras.models import load_model, Model\n",
    "from keras.layers import Input, Dense, LSTM, Activation, Masking, Dropout, Conv1D, GlobalAveragePooling1D, MaxPooling1D, Add\n",
    "from keras.callbacks import History\n",
    "from keras import callbacks\n",
    "\n",
    "# define path to save model\n",
    "model_path = './cl-fd3/binary_model.h5'\n",
    "path = './cl-fd3/'\n",
    "\n",
    "# read training data - It is the aircraft engine run-to-failure data.\n",
    "train_df = pd.read_csv('CMAPSSData/train_FD003.txt', sep=\" \", header=None)\n",
    "train_df.drop(train_df.columns[[4,5,9, 10,14,18,20,22, 23, 26, 27]], axis=1, inplace=True)\n",
    "\n",
    "train_df.columns = ['id', 'cycle', 'setting1', 'setting2', 's2', 's3','s4', 's7', 's8', 's9', 's11', 's12', 's13',\n",
    "                     's15', 's17', 's20', 's21']\n",
    "\n",
    "train_df = train_df.sort_values(['id','cycle'])\n",
    "\n",
    "# read test data - It is the aircraft engine operating data without failure events recorded.\n",
    "test_df = pd.read_csv('CMAPSSData/test_FD003.txt', sep=\" \", header=None)\n",
    "test_df.drop(test_df.columns[[4, 5, 9, 10, 14, 18, 20, 22, 23, 26, 27]], axis=1, inplace=True)\n",
    "test_df.columns = ['id', 'cycle', 'setting1', 'setting2', 's2', 's3','s4', 's7', 's8', 's9', 's11', 's12', 's13',\n",
    "                     's15', 's17', 's20', 's21']\n",
    "\n",
    "truth_df = pd.read_csv('CMAPSSData/RUL_FD003.txt', sep=\" \", header=None)\n",
    "truth_df.drop(truth_df.columns[[1]], axis=1, inplace=True)\n",
    "\n",
    "\n",
    "# Data Labeling - generate column RUL(Remaining Usefull Life or Time to Failure)\n",
    "rul = pd.DataFrame(train_df.groupby('id')['cycle'].max()).reset_index()\n",
    "rul.columns = ['id', 'max']\n",
    "train_df = train_df.merge(rul, on=['id'], how='left')\n",
    "train_df['RUL'] = train_df['max'] - train_df['cycle']\n",
    "train_df.drop('max', axis=1, inplace=True)\n",
    "# generate label columns for training data\n",
    "# we will only make use of \"label1\" for binary classification, \n",
    "# while trying to answer the question: is a specific engine going to fail within w1 cycles?\n",
    "w1 = 50\n",
    "w0 = 25\n",
    "train_df['label1'] = np.where(train_df['RUL'] <= w1, 1, 0 )\n",
    "train_df['label2'] = train_df['label1']\n",
    "train_df.loc[train_df['RUL'] <= w0, 'label2'] = 2\n",
    "\n",
    "# MinMax normalization (from 0 to 1)\n",
    "train_df['cycle_norm'] = train_df['cycle']\n",
    "cols_normalize = train_df.columns.difference(['id','cycle','RUL','label1','label2'])\n",
    "min_max_scaler = preprocessing.MinMaxScaler()\n",
    "norm_train_df = pd.DataFrame(min_max_scaler.fit_transform(train_df[cols_normalize]), \n",
    "                             columns=cols_normalize, \n",
    "                             index=train_df.index)\n",
    "join_df = train_df[train_df.columns.difference(cols_normalize)].join(norm_train_df)\n",
    "train_df = join_df.reindex(columns = train_df.columns)\n",
    "\n",
    "# MinMax normalization (from 0 to 1)\n",
    "test_df['cycle_norm'] = test_df['cycle']\n",
    "norm_test_df = pd.DataFrame(min_max_scaler.transform(test_df[cols_normalize]), \n",
    "                            columns=cols_normalize, \n",
    "                            index=test_df.index)\n",
    "test_join_df = test_df[test_df.columns.difference(cols_normalize)].join(norm_test_df)\n",
    "test_df = test_join_df.reindex(columns = test_df.columns)\n",
    "test_df = test_df.reset_index(drop=True)\n",
    "\n",
    "# We use the ground truth dataset to generate labels for the test data.\n",
    "# generate column max for test data\n",
    "rul = pd.DataFrame(test_df.groupby('id')['cycle'].max()).reset_index()\n",
    "rul.columns = ['id', 'max']\n",
    "truth_df.columns = ['more']\n",
    "truth_df['id'] = truth_df.index + 1\n",
    "truth_df['max'] = rul['max'] + truth_df['more']\n",
    "truth_df.drop('more', axis=1, inplace=True)\n",
    "\n",
    "# generate RUL for test data\n",
    "test_df = test_df.merge(truth_df, on=['id'], how='left')\n",
    "test_df['RUL'] = test_df['max'] - test_df['cycle']\n",
    "test_df.drop('max', axis=1, inplace=True)\n",
    "\n",
    "# generate label columns w0 and w1 for test data\n",
    "test_df['label1'] = np.where(test_df['RUL'] <= w1, 1, 0 )\n",
    "test_df['label2'] = test_df['label1']\n",
    "test_df.loc[test_df['RUL'] <= w0, 'label2'] = 2\n",
    "\n",
    "# pick a large window size of 50 cycles\n",
    "sequence_length = 50\n",
    "\n",
    "# function to reshape features into (samples, time steps, features) \n",
    "def gen_sequence(id_df, seq_length, seq_cols):\n",
    "    \"\"\" Only sequences that meet the window-length are considered, no padding is used. This means for testing\n",
    "    we need to drop those which are below the window-length. An alternative would be to pad sequences so that\n",
    "    we can use shorter ones \"\"\"\n",
    "    # for one id I put all the rows in a single matrix\n",
    "    data_matrix = id_df[seq_cols].values\n",
    "    num_elements = data_matrix.shape[0]\n",
    "    for start, stop in zip(range(0, num_elements-seq_length), range(seq_length, num_elements)):\n",
    "        yield data_matrix[start:stop, :]\n",
    "        \n",
    "# pick the feature columns \n",
    "sensor_cols = ['s2', 's3','s4', 's7', 's8', 's9', 's11', 's12', 's13','s15', 's17', 's20', 's21']\n",
    "sequence_cols = ['setting1', 'setting2', 'cycle_norm']\n",
    "sequence_cols.extend(sensor_cols)\n",
    "\n",
    "# generator for the sequences\n",
    "seq_gen = (list(gen_sequence(train_df[train_df['id']==id], sequence_length, sequence_cols)) \n",
    "           for id in train_df['id'].unique())\n",
    "\n",
    "# generate sequences and convert to numpy array\n",
    "seq_array = np.concatenate(list(seq_gen)).astype(np.float32)\n",
    "seq_array.shape\n",
    "\n",
    "# function to generate labels\n",
    "def gen_labels(id_df, seq_length, label):\n",
    "    data_matrix = id_df[label].values\n",
    "    num_elements = data_matrix.shape[0]\n",
    "    return data_matrix[seq_length:num_elements, :]\n",
    "\n",
    "# generate labels\n",
    "label_gen = [gen_labels(train_df[train_df['id']==id], sequence_length, ['label1']) \n",
    "             for id in train_df['id'].unique()]\n",
    "label_array = np.concatenate(label_gen).astype(np.float32)\n",
    "label_array.shape\n",
    "\n",
    "nb_features = seq_array.shape[2]\n",
    "nb_out = label_array.shape[1]\n",
    "\n",
    "history = History()\n",
    "\n",
    "input_layer = Input(shape=(sequence_length, nb_features))\n",
    "x = Conv1D(50, 6, activation = 'relu')(input_layer)\n",
    "a = Conv1D(50, 6, activation = 'relu')(x)\n",
    "x = Conv1D(50, 6, activation = 'relu')(x)\n",
    "b = Add()([a,x]) #SKIP CONNECTION ADDED\n",
    "x = MaxPooling1D(3)(b)\n",
    "x = Conv1D(100, 6, activation = 'relu')(x)\n",
    "x = Conv1D(100, 6, activation = 'relu')(x)\n",
    "x = MaxPooling1D(3)(x)\n",
    "x = Dropout(0.2)(x)\n",
    "x = LSTM( units=50, input_shape=(1,100), return_sequences=True)(x)\n",
    "x = LSTM( units=100, return_sequences=True)(x)\n",
    "c = LSTM( units=100, return_sequences=True)(x)\n",
    "x = LSTM( units=100, return_sequences=True)(x)\n",
    "d = Add()([c,x]) #skip connection\n",
    "x = GlobalAveragePooling1D()(x)\n",
    "x = Dropout(0.2)(x)\n",
    "output_layer = Dense(units=nb_out, activation ='sigmoid')(x)\n",
    "model = Model(input_layer, output_layer)\n",
    "model.compile(loss=\"binary_crossentropy\", optimizer=\"adam\", metrics=['accuracy', keras.metrics.Precision(), keras.metrics.Recall()])\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "593d4dec",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(19720, 50, 16)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seq_array.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "66a20fbb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    16596.000000\n",
       "mean         0.046939\n",
       "std          0.211515\n",
       "min          0.000000\n",
       "25%          0.000000\n",
       "50%          0.000000\n",
       "75%          0.000000\n",
       "max          1.000000\n",
       "Name: label1, dtype: float64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_df['label1'].describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "1dd5c654",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[0.54651165, 0.23076923, 0.0019084 , 0.38875878, 0.39909962,\n",
       "        0.30936038, 0.23659004, 0.64705884, 0.20243102, 0.30857143,\n",
       "        0.23688233, 0.6547619 , 0.6856591 , 0.36363637, 0.48809522,\n",
       "        0.53483623],\n",
       "       [0.41860464, 0.30769232, 0.00381679, 0.31381732, 0.3532981 ,\n",
       "        0.44539782, 0.23084292, 0.6647059 , 0.2414845 , 0.30285713,\n",
       "        0.21701477, 0.6369048 , 0.5644616 , 0.27272728, 0.4047619 ,\n",
       "        0.45857686],\n",
       "       [0.38372093, 0.53846157, 0.00572519, 0.48711944, 0.41710708,\n",
       "        0.2372855 , 0.26819924, 0.64705884, 0.2153256 , 0.31428573,\n",
       "        0.2404483 , 0.6845238 , 0.55890876, 0.36363637, 0.4702381 ,\n",
       "        0.39196587],\n",
       "       [0.59302324, 0.46153846, 0.00763359, 0.19672132, 0.47621843,\n",
       "        0.32121685, 0.24568966, 0.67058825, 0.26879883, 0.26285714,\n",
       "        0.24503312, 0.6547619 , 0.5567359 , 0.36363637, 0.57738096,\n",
       "        0.5433714 ],\n",
       "       [0.56395346, 0.07692308, 0.00954198, 0.32786885, 0.38735566,\n",
       "        0.3580343 , 0.24616858, 0.64705884, 0.26343763, 0.21714285,\n",
       "        0.23688233, 0.64880955, 0.6250604 , 0.45454547, 0.44642857,\n",
       "        0.5153539 ],\n",
       "       [0.27906978, 0.61538464, 0.01145038, 0.40749416, 0.41710708,\n",
       "        0.374883  , 0.23802681, 0.67058825, 0.26191247, 0.3942857 ,\n",
       "        0.23892002, 0.64880955, 0.5502173 , 0.27272728, 0.39880952,\n",
       "        0.49735597],\n",
       "       [0.45930234, 0.07692308, 0.01335878, 0.34660423, 0.4701507 ,\n",
       "        0.3123245 , 0.25095785, 0.68235296, 0.21056524, 0.35428572,\n",
       "        0.22058074, 0.61904764, 0.63423467, 0.27272728, 0.52380955,\n",
       "        0.5440208 ],\n",
       "       [0.3488372 , 0.46153846, 0.01526718, 0.18735363, 0.46016833,\n",
       "        0.45943838, 0.2164751 , 0.67058825, 0.22748071, 0.35428572,\n",
       "        0.20682628, 0.64285713, 0.61033314, 0.36363637, 0.48809522,\n",
       "        0.5535764 ],\n",
       "       [0.6104651 , 0.30769232, 0.01717557, 0.37002343, 0.45194754,\n",
       "        0.3929797 , 0.24760537, 0.64117646, 0.24060637, 0.37714285,\n",
       "        0.21344881, 0.63095236, 0.5852245 , 0.27272728, 0.45833334,\n",
       "        0.5617404 ],\n",
       "       [0.6511628 , 0.30769232, 0.01908397, 0.33723652, 0.3538853 ,\n",
       "        0.23120125, 0.22461686, 0.65294117, 0.23228729, 0.37714285,\n",
       "        0.22720326, 0.6547619 , 0.645099  , 0.36363637, 0.47619048,\n",
       "        0.50106686],\n",
       "       [0.29069766, 0.15384616, 0.02099237, 0.24355972, 0.37071833,\n",
       "        0.3226209 , 0.22701149, 0.67058825, 0.2647317 , 0.3257143 ,\n",
       "        0.22210902, 0.625     , 0.4893771 , 0.36363637, 0.54761904,\n",
       "        0.45774192],\n",
       "       [0.4360465 , 0.53846157, 0.02290076, 0.2529274 , 0.4967704 ,\n",
       "        0.28845555, 0.23515326, 0.63529414, 0.22438416, 0.25714287,\n",
       "        0.21854305, 0.64880955, 0.5451473 , 0.27272728, 0.51785713,\n",
       "        0.5146117 ],\n",
       "       [0.5523256 , 0.53846157, 0.02480916, 0.4262295 , 0.4499902 ,\n",
       "        0.3719189 , 0.2662835 , 0.65882355, 0.22812775, 0.33714285,\n",
       "        0.24299541, 0.61904764, 0.655239  , 0.18181819, 0.5119048 ,\n",
       "        0.4762965 ],\n",
       "       [0.20930232, 0.7692308 , 0.02671756, 0.35128805, 0.32295948,\n",
       "        0.30327612, 0.20833333, 0.61764705, 0.2399131 , 0.34285715,\n",
       "        0.25216505, 0.6369048 , 0.59005314, 0.18181819, 0.50595236,\n",
       "        0.5408665 ],\n",
       "       [0.6395349 , 0.6923077 , 0.02862595, 0.19672132, 0.4505774 ,\n",
       "        0.2875195 , 0.20162836, 0.65882355, 0.24569026, 0.25714287,\n",
       "        0.22975038, 0.625     , 0.6359247 , 0.27272728, 0.4702381 ,\n",
       "        0.46302995],\n",
       "       [0.3895349 , 0.53846157, 0.03053435, 0.24824356, 0.524956  ,\n",
       "        0.28112325, 0.26101533, 0.64705884, 0.19762444, 0.17714286,\n",
       "        0.21446764, 0.5952381 , 0.5306615 , 0.45454547, 0.51785713,\n",
       "        0.46312273],\n",
       "       [0.29069766, 0.61538464, 0.03244275, 0.36299765, 0.2869446 ,\n",
       "        0.39282373, 0.22030652, 0.62941176, 0.26755095, 0.29142857,\n",
       "        0.22567499, 0.64285713, 0.6378561 , 0.36363637, 0.5654762 ,\n",
       "        0.62362   ],\n",
       "       [0.44186047, 0.46153846, 0.03435114, 0.34192038, 0.38637698,\n",
       "        0.3399376 , 0.23850575, 0.63529414, 0.26080325, 0.26857144,\n",
       "        0.2175242 , 0.625     , 0.52245295, 0.27272728, 0.5       ,\n",
       "        0.49670655],\n",
       "       [0.3372093 , 0.7692308 , 0.03625954, 0.3325527 , 0.30045018,\n",
       "        0.28907958, 0.23371647, 0.62941176, 0.2463373 , 0.26857144,\n",
       "        0.26744777, 0.6547619 , 0.6492033 , 0.27272728, 0.41666666,\n",
       "        0.5730587 ],\n",
       "       [0.50581396, 0.30769232, 0.03816794, 0.41686183, 0.50655705,\n",
       "        0.32683307, 0.21024904, 0.64117646, 0.2619587 , 0.3257143 ,\n",
       "        0.25471216, 0.64285713, 0.6414775 , 0.27272728, 0.44642857,\n",
       "        0.5902217 ],\n",
       "       [0.56395346, 0.7692308 , 0.04007633, 0.30210772, 0.42885104,\n",
       "        0.31216848, 0.23467433, 0.63529414, 0.26685768, 0.36      ,\n",
       "        0.25012735, 0.6547619 , 0.6207146 , 0.36363637, 0.50595236,\n",
       "        0.44595975],\n",
       "       [0.16860466, 0.3846154 , 0.04198473, 0.28337237, 0.42532787,\n",
       "        0.26443058, 0.22749043, 0.64117646, 0.23325785, 0.2742857 ,\n",
       "        0.20631686, 0.6547619 , 0.5753259 , 0.27272728, 0.5297619 ,\n",
       "        0.53344464],\n",
       "       [0.8313953 , 0.61538464, 0.04389313, 0.2763466 , 0.4077119 ,\n",
       "        0.27691108, 0.22892721, 0.62352943, 0.252022  , 0.41142857,\n",
       "        0.24554254, 0.64880955, 0.62747467, 0.27272728, 0.53571427,\n",
       "        0.54940164],\n",
       "       [0.44767442, 0.15384616, 0.04580153, 0.25526932, 0.44079077,\n",
       "        0.46115446, 0.20114942, 0.64117646, 0.26861393, 0.35428572,\n",
       "        0.21956189, 0.66071427, 0.5767745 , 0.27272728, 0.4047619 ,\n",
       "        0.4746266 ],\n",
       "       [0.5523256 , 0.3846154 , 0.04770992, 0.35831383, 0.41025642,\n",
       "        0.33868954, 0.24808429, 0.63529414, 0.21902297, 0.33714285,\n",
       "        0.21854305, 0.64880955, 0.5731531 , 0.36363637, 0.45238096,\n",
       "        0.45969012],\n",
       "       [0.5290698 , 0.3846154 , 0.04961832, 0.2997658 , 0.3615189 ,\n",
       "        0.36037442, 0.21886973, 0.63529414, 0.21828349, 0.36571428,\n",
       "        0.21956189, 0.6547619 , 0.62023175, 0.18181819, 0.5297619 ,\n",
       "        0.46729752],\n",
       "       [0.4651163 , 0.7692308 , 0.05152672, 0.5081967 , 0.27598357,\n",
       "        0.36162245, 0.24185824, 0.64117646, 0.17483939, 0.2742857 ,\n",
       "        0.22516556, 0.66071427, 0.6470304 , 0.27272728, 0.50595236,\n",
       "        0.5510715 ],\n",
       "       [0.43023255, 0.3846154 , 0.05343511, 0.28103045, 0.4541006 ,\n",
       "        0.24945398, 0.21503831, 0.62941176, 0.23200998, 0.30285713,\n",
       "        0.20580743, 0.64285713, 0.5886045 , 0.45454547, 0.51785713,\n",
       "        0.5391966 ],\n",
       "       [0.40697673, 0.15384616, 0.05534351, 0.39578456, 0.44411823,\n",
       "        0.30592823, 0.24760537, 0.6647059 , 0.23543005, 0.37142858,\n",
       "        0.23841059, 0.61904764, 0.7122163 , 0.36363637, 0.5654762 ,\n",
       "        0.49884033],\n",
       "       [0.47674417, 0.53846157, 0.05725191, 0.34660423, 0.4272852 ,\n",
       "        0.31840873, 0.19636016, 0.61764705, 0.21181309, 0.31428573,\n",
       "        0.23331635, 0.64285713, 0.53186864, 0.18181819, 0.6130952 ,\n",
       "        0.51730216],\n",
       "       [0.60465115, 0.07692308, 0.05916031, 0.2974239 , 0.45801526,\n",
       "        0.2399376 , 0.27155173, 0.6647059 , 0.22262791, 0.2742857 ,\n",
       "        0.25216505, 0.6130952 , 0.62578464, 0.27272728, 0.50595236,\n",
       "        0.45505148],\n",
       "       [0.6511628 , 0.23076923, 0.0610687 , 0.38641685, 0.29829714,\n",
       "        0.35694227, 0.21982759, 0.64705884, 0.1934649 , 0.33142856,\n",
       "        0.21293938, 0.6369048 , 0.6187832 , 0.36363637, 0.58928573,\n",
       "        0.49531496],\n",
       "       [0.3604651 , 0.30769232, 0.0629771 , 0.3208431 , 0.43296143,\n",
       "        0.41107646, 0.22365901, 0.62352943, 0.23131673, 0.4457143 ,\n",
       "        0.21701477, 0.64880955, 0.64099467, 0.27272728, 0.50595236,\n",
       "        0.56229705],\n",
       "       [0.18023255, 0.6923077 , 0.0648855 , 0.22482435, 0.5175181 ,\n",
       "        0.300624  , 0.19827586, 0.6647059 , 0.23898877, 0.49142858,\n",
       "        0.22007132, 0.64880955, 0.65306616, 0.36363637, 0.43452382,\n",
       "        0.3866778 ],\n",
       "       [0.5697674 , 0.15384616, 0.0667939 , 0.37704918, 0.34370717,\n",
       "        0.29375976, 0.25095785, 0.65882355, 0.22798909, 0.36      ,\n",
       "        0.22669384, 0.6547619 , 0.62240463, 0.45454547, 0.44047618,\n",
       "        0.6096113 ],\n",
       "       [0.26744187, 0.3846154 , 0.06870229, 0.2857143 , 0.48326483,\n",
       "        0.31076443, 0.24952108, 0.61764705, 0.25682858, 0.36571428,\n",
       "        0.24503312, 0.64880955, 0.42056978, 0.45454547, 0.5119048 ,\n",
       "        0.53928936],\n",
       "       [0.73255813, 0.15384616, 0.07061069, 0.38641685, 0.4777843 ,\n",
       "        0.3823713 , 0.25957856, 0.65294117, 0.24860193, 0.34857142,\n",
       "        0.2470708 , 0.625     , 0.52028006, 0.36363637, 0.39285713,\n",
       "        0.5126635 ],\n",
       "       [0.877907  , 0.46153846, 0.07251909, 0.17798595, 0.28361714,\n",
       "        0.3653666 , 0.22509578, 0.65882355, 0.26126543, 0.2857143 ,\n",
       "        0.22873153, 0.63095236, 0.5895703 , 0.36363637, 0.44047618,\n",
       "        0.44688746],\n",
       "       [0.5406977 , 0.30769232, 0.07442748, 0.323185  , 0.3836367 ,\n",
       "        0.28393134, 0.23467433, 0.62352943, 0.2393585 , 0.31428573,\n",
       "        0.23942944, 0.64880955, 0.57580876, 0.54545456, 0.45833334,\n",
       "        0.49698487],\n",
       "       [0.41860464, 0.30769232, 0.07633588, 0.26229507, 0.35701704,\n",
       "        0.36911076, 0.2394636 , 0.62941176, 0.2522531 , 0.36      ,\n",
       "        0.23841059, 0.61904764, 0.4818928 , 0.36363637, 0.53571427,\n",
       "        0.41868448],\n",
       "       [0.60465115, 0.30769232, 0.07824428, 0.2857143 , 0.3994911 ,\n",
       "        0.40795633, 0.22509578, 0.64705884, 0.24818598, 0.3257143 ,\n",
       "        0.22567499, 0.61904764, 0.62409467, 0.36363637, 0.4107143 ,\n",
       "        0.524724  ],\n",
       "       [0.5523256 , 0.7692308 , 0.08015267, 0.4379391 , 0.33646506,\n",
       "        0.34820592, 0.20833333, 0.62352943, 0.25206822, 0.23428571,\n",
       "        0.23688233, 0.625     , 0.6673105 , 0.36363637, 0.57738096,\n",
       "        0.5224047 ],\n",
       "       [0.61627907, 0.53846157, 0.08206107, 0.43325526, 0.41573694,\n",
       "        0.47737908, 0.15517241, 0.64117646, 0.23164025, 0.3257143 ,\n",
       "        0.2552216 , 0.60119045, 0.6690005 , 0.09090909, 0.5952381 ,\n",
       "        0.44428983],\n",
       "       [0.6104651 , 0.53846157, 0.08396947, 0.4028103 , 0.39009592,\n",
       "        0.48190328, 0.22126436, 0.62352943, 0.22646393, 0.2742857 ,\n",
       "        0.24401426, 0.625     , 0.67117333, 0.27272728, 0.5       ,\n",
       "        0.408665  ],\n",
       "       [0.2848837 , 0.6923077 , 0.08587787, 0.2763466 , 0.3640634 ,\n",
       "        0.36567864, 0.2394636 , 0.68235296, 0.24624486, 0.29714286,\n",
       "        0.2175242 , 0.66071427, 0.65306616, 0.36363637, 0.42261904,\n",
       "        0.58289266],\n",
       "       [0.3895349 , 0.6923077 , 0.08778626, 0.41217798, 0.5441378 ,\n",
       "        0.33073324, 0.21551724, 0.65294117, 0.25756806, 0.33714285,\n",
       "        0.21701477, 0.61904764, 0.52921295, 0.36363637, 0.57738096,\n",
       "        0.3677521 ],\n",
       "       [0.51744187, 0.46153846, 0.08969466, 0.37470725, 0.37130553,\n",
       "        0.34492978, 0.21791188, 0.65882355, 0.2264177 , 0.33142856,\n",
       "        0.2302598 , 0.66071427, 0.58788025, 0.54545456, 0.50595236,\n",
       "        0.41395307],\n",
       "       [0.6744186 , 0.15384616, 0.09160306, 0.35831383, 0.43942064,\n",
       "        0.26162246, 0.2059387 , 0.67058825, 0.2689837 , 0.33714285,\n",
       "        0.21242996, 0.64285713, 0.541043  , 0.18181819, 0.44642857,\n",
       "        0.47277114],\n",
       "       [0.45348838, 0.7692308 , 0.09351145, 0.3676815 , 0.5081229 ,\n",
       "        0.33463338, 0.22318007, 0.63529414, 0.23362759, 0.22285715,\n",
       "        0.21599592, 0.66071427, 0.6059874 , 0.36363637, 0.47619048,\n",
       "        0.46590593],\n",
       "       [0.43023255, 0.46153846, 0.09541985, 0.31850117, 0.458211  ,\n",
       "        0.20592824, 0.24425288, 0.64117646, 0.22928317, 0.29714286,\n",
       "        0.23586348, 0.625     , 0.48382425, 0.27272728, 0.43452382,\n",
       "        0.56201875]], dtype=float32)"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "seq_array[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "49dff239",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "188/188 [==============================] - 9s 26ms/step - loss: 0.4545 - accuracy: 0.8003 - precision: 0.7726 - recall: 0.3226 - val_loss: 0.2289 - val_accuracy: 0.9006 - val_precision: 0.7461 - val_recall: 0.9333\n",
      "Epoch 2/100\n",
      "  7/188 [>.............................] - ETA: 3s - loss: 0.2006 - accuracy: 0.9271 - precision: 0.8636 - recall: 0.8769"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/shrutirajvanshsingh/opt/anaconda3/lib/python3.8/site-packages/keras/engine/functional.py:1410: CustomMaskWarning: Custom mask layers require a config and must override get_config. When loading, the custom mask layer must be passed to the custom_objects argument.\n",
      "  layer_config = serialize_layer_fn(layer)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "188/188 [==============================] - 4s 20ms/step - loss: 0.1569 - accuracy: 0.9376 - precision: 0.8943 - recall: 0.8605 - val_loss: 0.1957 - val_accuracy: 0.9138 - val_precision: 0.9381 - val_recall: 0.7137\n",
      "Epoch 3/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.1240 - accuracy: 0.9486 - precision: 0.9118 - recall: 0.8873 - val_loss: 0.1306 - val_accuracy: 0.9432 - val_precision: 0.8872 - val_recall: 0.8941\n",
      "Epoch 4/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.1107 - accuracy: 0.9535 - precision: 0.9185 - recall: 0.9001 - val_loss: 0.1312 - val_accuracy: 0.9239 - val_precision: 0.8409 - val_recall: 0.8706\n",
      "Epoch 5/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0985 - accuracy: 0.9598 - precision: 0.9268 - recall: 0.9170 - val_loss: 0.1275 - val_accuracy: 0.9331 - val_precision: 0.8363 - val_recall: 0.9216\n",
      "Epoch 6/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0897 - accuracy: 0.9629 - precision: 0.9328 - recall: 0.9230 - val_loss: 0.1067 - val_accuracy: 0.9493 - val_precision: 0.9325 - val_recall: 0.8667\n",
      "Epoch 7/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0897 - accuracy: 0.9643 - precision: 0.9357 - recall: 0.9255 - val_loss: 0.1003 - val_accuracy: 0.9615 - val_precision: 0.8834 - val_recall: 0.9804\n",
      "Epoch 8/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0758 - accuracy: 0.9696 - precision: 0.9423 - recall: 0.9399 - val_loss: 0.1149 - val_accuracy: 0.9402 - val_precision: 0.9495 - val_recall: 0.8118\n",
      "Epoch 9/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0739 - accuracy: 0.9712 - precision: 0.9487 - recall: 0.9395 - val_loss: 0.1116 - val_accuracy: 0.9422 - val_precision: 0.9541 - val_recall: 0.8157\n",
      "Epoch 10/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0594 - accuracy: 0.9759 - precision: 0.9571 - recall: 0.9494 - val_loss: 0.0827 - val_accuracy: 0.9635 - val_precision: 0.9469 - val_recall: 0.9098\n",
      "Epoch 11/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0637 - accuracy: 0.9736 - precision: 0.9498 - recall: 0.9482 - val_loss: 0.0792 - val_accuracy: 0.9594 - val_precision: 0.9151 - val_recall: 0.9294\n",
      "Epoch 12/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0579 - accuracy: 0.9770 - precision: 0.9545 - recall: 0.9569 - val_loss: 0.1471 - val_accuracy: 0.9371 - val_precision: 1.0000 - val_recall: 0.7569\n",
      "Epoch 13/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0592 - accuracy: 0.9763 - precision: 0.9589 - recall: 0.9490 - val_loss: 0.0885 - val_accuracy: 0.9513 - val_precision: 0.9559 - val_recall: 0.8510\n",
      "Epoch 14/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0627 - accuracy: 0.9748 - precision: 0.9532 - recall: 0.9490 - val_loss: 0.0567 - val_accuracy: 0.9746 - val_precision: 0.9713 - val_recall: 0.9294\n",
      "Epoch 15/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0469 - accuracy: 0.9810 - precision: 0.9665 - recall: 0.9598 - val_loss: 0.0779 - val_accuracy: 0.9513 - val_precision: 0.9600 - val_recall: 0.8471\n",
      "Epoch 16/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0504 - accuracy: 0.9791 - precision: 0.9624 - recall: 0.9567 - val_loss: 0.0526 - val_accuracy: 0.9797 - val_precision: 0.9681 - val_recall: 0.9529\n",
      "Epoch 17/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0407 - accuracy: 0.9847 - precision: 0.9715 - recall: 0.9692 - val_loss: 0.0990 - val_accuracy: 0.9584 - val_precision: 1.0000 - val_recall: 0.8392\n",
      "Epoch 18/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0388 - accuracy: 0.9850 - precision: 0.9729 - recall: 0.9690 - val_loss: 0.0449 - val_accuracy: 0.9797 - val_precision: 0.9537 - val_recall: 0.9686\n",
      "Epoch 19/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0356 - accuracy: 0.9854 - precision: 0.9733 - recall: 0.9703 - val_loss: 0.0541 - val_accuracy: 0.9807 - val_precision: 0.9574 - val_recall: 0.9686\n",
      "Epoch 20/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0328 - accuracy: 0.9867 - precision: 0.9746 - recall: 0.9738 - val_loss: 0.0539 - val_accuracy: 0.9746 - val_precision: 0.9600 - val_recall: 0.9412\n",
      "Epoch 21/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0328 - accuracy: 0.9872 - precision: 0.9752 - recall: 0.9752 - val_loss: 0.0509 - val_accuracy: 0.9807 - val_precision: 0.9538 - val_recall: 0.9725\n",
      "Epoch 22/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0300 - accuracy: 0.9883 - precision: 0.9787 - recall: 0.9761 - val_loss: 0.0560 - val_accuracy: 0.9787 - val_precision: 0.9432 - val_recall: 0.9765\n",
      "Epoch 23/100\n",
      "188/188 [==============================] - 4s 19ms/step - loss: 0.0283 - accuracy: 0.9890 - precision: 0.9801 - recall: 0.9773 - val_loss: 0.1534 - val_accuracy: 0.9483 - val_precision: 1.0000 - val_recall: 0.8000\n",
      "Epoch 24/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0285 - accuracy: 0.9894 - precision: 0.9806 - recall: 0.9783 - val_loss: 0.1513 - val_accuracy: 0.9473 - val_precision: 1.0000 - val_recall: 0.7961\n",
      "Epoch 25/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0339 - accuracy: 0.9867 - precision: 0.9742 - recall: 0.9744 - val_loss: 0.0747 - val_accuracy: 0.9726 - val_precision: 0.9254 - val_recall: 0.9725\n",
      "Epoch 26/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0301 - accuracy: 0.9885 - precision: 0.9793 - recall: 0.9763 - val_loss: 0.0777 - val_accuracy: 0.9686 - val_precision: 0.9029 - val_recall: 0.9843\n",
      "Epoch 27/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0272 - accuracy: 0.9891 - precision: 0.9809 - recall: 0.9769 - val_loss: 0.0504 - val_accuracy: 0.9828 - val_precision: 0.9798 - val_recall: 0.9529\n",
      "Epoch 28/100\n",
      "188/188 [==============================] - 4s 20ms/step - loss: 0.0258 - accuracy: 0.9904 - precision: 0.9812 - recall: 0.9816 - val_loss: 0.0999 - val_accuracy: 0.9686 - val_precision: 0.9480 - val_recall: 0.9294\n",
      "dict_keys(['loss', 'accuracy', 'precision', 'recall', 'val_loss', 'val_accuracy', 'val_precision', 'val_recall'])\n"
     ]
    }
   ],
   "source": [
    "# fit the network\n",
    "history = model.fit(seq_array, label_array, epochs=100, batch_size=100, validation_split=0.05, verbose=1,\n",
    "          callbacks = [keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0, patience=10, verbose=0, mode='min'),\n",
    "                       keras.callbacks.ModelCheckpoint(model_path,monitor='val_loss', save_best_only=True, mode='min', verbose=0)]\n",
    "          )\n",
    "\n",
    "# list all data in history\n",
    "print(history.history.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "47932502",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for Accuracy\n",
    "fig_acc = plt.figure(figsize=(10, 10))\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "fig_acc.savefig(path+\"/model_accuracy.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "df6c8325",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for Loss\n",
    "fig_acc = plt.figure(figsize=(10, 10))\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', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "fig_acc.savefig(path+\"/model_loss.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "85d4a1e3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "99/99 [==============================] - 1s 8ms/step - loss: 0.0424 - accuracy: 0.9826 - precision: 0.9482 - recall: 0.9867\n",
      "Accurracy: 0.9826064705848694\n",
      "Precision: 0.9481816291809082\n",
      "Recall: 0.9866666793823242\n"
     ]
    }
   ],
   "source": [
    "# training metrics\n",
    "scores = model.evaluate(seq_array, label_array, verbose=1, batch_size=200)\n",
    "print('Accurracy: {}'.format(scores[1]))\n",
    "print('Precision: {}'.format(scores[2]))\n",
    "print('Recall: {}'.format(scores[3]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "9c91568d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "617/617 [==============================] - 2s 2ms/step\n"
     ]
    }
   ],
   "source": [
    "# make predictions and compute confusion matrix\n",
    "predict_y = model.predict(seq_array,verbose=1) \n",
    "y_pred=np.argmax(predict_y, axis=1)\n",
    "y_true = label_array\n",
    "test_set = pd.DataFrame(y_pred)\n",
    "test_set.to_csv(path+'/binary_submit_train.csv', index = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "0cdfbbae",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "seq_array_test_last\n",
      "(97, 50, 16)\n",
      "y_mask\n",
      "[True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, True, False, True, True, True, True, True, True, True, True, True, True, True, True]\n",
      "(97, 1)\n",
      "label_array_test_last\n",
      "4/4 - 1s - loss: 0.2816 - accuracy: 0.9381 - precision: 0.8286 - recall: 1.0000 - 1s/epoch - 265ms/step\n",
      "Accurracy: 0.938144326210022\n"
     ]
    }
   ],
   "source": [
    "# We pick the last sequence for each id in the test data\n",
    "from sklearn.metrics import f1_score\n",
    "seq_array_test_last = [test_df[test_df['id']==id][sequence_cols].values[-sequence_length:] \n",
    "                       for id in test_df['id'].unique() if len(test_df[test_df['id']==id]) >= sequence_length]\n",
    "\n",
    "seq_array_test_last = np.asarray(seq_array_test_last).astype(np.float32)\n",
    "print(\"seq_array_test_last\")\n",
    "#print(seq_array_test_last)\n",
    "print(seq_array_test_last.shape)\n",
    "\n",
    "# Similarly, we pick the labels\n",
    "\n",
    "#print(\"y_mask\")\n",
    "# serve per prendere solo le label delle sequenze che sono almeno lunghe 50\n",
    "y_mask = [len(test_df[test_df['id']==id]) >= sequence_length for id in test_df['id'].unique()]\n",
    "print(\"y_mask\")\n",
    "print(y_mask)\n",
    "label_array_test_last = test_df.groupby('id')['label1'].nth(-1)[y_mask].values\n",
    "label_array_test_last = label_array_test_last.reshape(label_array_test_last.shape[0],1).astype(np.float32)\n",
    "print(label_array_test_last.shape)\n",
    "print(\"label_array_test_last\")\n",
    "#print(label_array_test_last)\n",
    "\n",
    "# if best iteration's model was saved then load and use it\n",
    "if os.path.isfile(model_path):\n",
    "    estimator = load_model(model_path)\n",
    "\n",
    "# test metrics\n",
    "scores_test = estimator.evaluate(seq_array_test_last, label_array_test_last, verbose=2)\n",
    "print('Accurracy: {}'.format(scores_test[1]))\n",
    "\n",
    "# make predictions and compute confusion matrix\n",
    "predict_y_test = estimator.predict(seq_array_test_last)\n",
    "y_pred_test = np.argmax(predict_y_test,axis=1)\n",
    "y_true_test = label_array_test_last\n",
    "\n",
    "test_set = pd.DataFrame(y_pred_test)\n",
    "test_set.to_csv(path+'binary_submit_test.csv', index = None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a9735537",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8d37c0ca",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ade70169",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a6602701",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f45d5e4f",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a120bf37",
   "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": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}