Bidirectional LSTM Model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import random \n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras import Input, Model\n",
    "from tensorflow.keras.layers import Bidirectional, LSTM, Dense, GlobalMaxPooling1D, concatenate, GlobalAveragePooling1D"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "db = pd.read_csv('NIFTY50_all.csv', header=0) #dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "         Date      Symbol Series  Prev Close    Open     High    Low   Last  \\\n",
      "0  2007-11-27  MUNDRAPORT     EQ      440.00  770.00  1050.00  770.0  959.0   \n",
      "1  2007-11-28  MUNDRAPORT     EQ      962.90  984.00   990.00  874.0  885.0   \n",
      "2  2007-11-29  MUNDRAPORT     EQ      893.90  909.00   914.75  841.0  887.0   \n",
      "3  2007-11-30  MUNDRAPORT     EQ      884.20  890.00   958.00  890.0  929.0   \n",
      "4  2007-12-03  MUNDRAPORT     EQ      921.55  939.75   995.00  922.0  980.0   \n",
      "\n",
      "    Close    VWAP    Volume      Turnover  Trades  Deliverable Volume  \\\n",
      "0  962.90  984.72  27294366  2.687719e+15     NaN           9859619.0   \n",
      "1  893.90  941.38   4581338  4.312765e+14     NaN           1453278.0   \n",
      "2  884.20  888.09   5124121  4.550658e+14     NaN           1069678.0   \n",
      "3  921.55  929.17   4609762  4.283257e+14     NaN           1260913.0   \n",
      "4  969.30  965.65   2977470  2.875200e+14     NaN            816123.0   \n",
      "\n",
      "   %Deliverble  \n",
      "0       0.3612  \n",
      "1       0.3172  \n",
      "2       0.2088  \n",
      "3       0.2735  \n",
      "4       0.2741  \n"
     ]
    }
   ],
   "source": [
    "print(db.head())\n",
    "\n",
    "x = {} #company set\n",
    "\n",
    "for i in range(len(db)):\n",
    "    x[db.Symbol[i]] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "min_return = min(db[['Close']].min(axis=0))\n",
    "max_return = max(db[['Close']].max(axis=0))\n",
    "\n",
    "db['Close'] = (db['Close'] - min_return) / (max_return - min_return) #normalisation (stock prices are now between 0-1)\n",
    "\n",
    "#db['Close'] = (db['Close'] * (max_return - min_return)) + min_return"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "65\n"
     ]
    }
   ],
   "source": [
    "print(len(x))\n",
    "\n",
    "#print(db.Date[:20])\n",
    "#print(db.Symbol[:20])\n",
    "\n",
    "sym = db.Symbol[0]\n",
    "\n",
    "temp = [[]]*65      #company wise matrix\n",
    "#print(temp)\n",
    "count = 0           #company number\n",
    "\n",
    "for i in range(len(db)):\n",
    "    if db.Symbol[i] == sym:\n",
    "        temp[count].append(db.Close[i])\n",
    "    else:\n",
    "        count += 1\n",
    "        sym = db.Symbol[i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "N = 10              #number of days\n",
    "n_rows = 10000      #10K rows ~ 11 mins of training\n",
    "X = []              #input\n",
    "Y = []              #output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(65):                         #company\n",
    "    for k in range(n_rows//65):             #fixed number of datapoints       \n",
    "        j = random.randint(0, len(temp[i])-N)           #choose starting point\n",
    "        X.append(temp[i][j:j+N])            #add 10 days from starting point to input\n",
    "        Y.append(temp[i][j+N])              #11th day is output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9945"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.33, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_model():\n",
    "    in_seq = Input(shape = (N, 1))\n",
    "\n",
    "    x = Bidirectional(LSTM(128, return_sequences=True))(in_seq)\n",
    "    x = Bidirectional(LSTM(128, return_sequences=True))(x)\n",
    "    x = Bidirectional(LSTM(64, return_sequences=True))(x) \n",
    "            \n",
    "    avg_pool = GlobalAveragePooling1D()(x)\n",
    "    max_pool = GlobalMaxPooling1D()(x)\n",
    "    conc = concatenate([avg_pool, max_pool])\n",
    "    conc = Dense(64, activation=\"relu\")(conc)\n",
    "    out = Dense(1, activation=\"linear\")(conc)      \n",
    "\n",
    "    model = Model(inputs=in_seq, outputs=out)\n",
    "    model.compile(loss=\"mse\", optimizer=\"adam\", metrics=['mae', 'mape'])    \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "4/4 [==============================] - 15s 857ms/step - loss: 0.0068 - mae: 0.0349 - mape: 170.3420\n",
      "Epoch 2/200\n",
      "4/4 [==============================] - 3s 774ms/step - loss: 0.0013 - mae: 0.0272 - mape: 414.1307\n",
      "Epoch 3/200\n",
      "4/4 [==============================] - 3s 763ms/step - loss: 9.5648e-04 - mae: 0.0251 - mape: 282.2142\n",
      "Epoch 4/200\n",
      "4/4 [==============================] - 3s 725ms/step - loss: 3.8800e-04 - mae: 0.0152 - mape: 252.9227\n",
      "Epoch 5/200\n",
      "4/4 [==============================] - 3s 729ms/step - loss: 2.3957e-04 - mae: 0.0088 - mape: 145.6939\n",
      "Epoch 6/200\n",
      "4/4 [==============================] - 3s 808ms/step - loss: 1.6626e-04 - mae: 0.0091 - mape: 66.7686\n",
      "Epoch 7/200\n",
      "4/4 [==============================] - 3s 789ms/step - loss: 5.8889e-05 - mae: 0.0049 - mape: 58.5237\n",
      "Epoch 8/200\n",
      "4/4 [==============================] - 3s 776ms/step - loss: 1.0244e-04 - mae: 0.0064 - mape: 107.7985\n",
      "Epoch 9/200\n",
      "4/4 [==============================] - 3s 704ms/step - loss: 5.9152e-05 - mae: 0.0041 - mape: 36.7082\n",
      "Epoch 10/200\n",
      "4/4 [==============================] - 3s 679ms/step - loss: 5.0289e-05 - mae: 0.0034 - mape: 31.1058\n",
      "Epoch 11/200\n",
      "4/4 [==============================] - 3s 850ms/step - loss: 5.0668e-05 - mae: 0.0037 - mape: 61.8426\n",
      "Epoch 12/200\n",
      "4/4 [==============================] - 4s 844ms/step - loss: 3.6554e-05 - mae: 0.0027 - mape: 30.1526\n",
      "Epoch 13/200\n",
      "4/4 [==============================] - 4s 828ms/step - loss: 4.8507e-05 - mae: 0.0024 - mape: 17.7606\n",
      "Epoch 14/200\n",
      "4/4 [==============================] - 3s 778ms/step - loss: 3.6316e-05 - mae: 0.0026 - mape: 40.1323\n",
      "Epoch 15/200\n",
      "4/4 [==============================] - 3s 737ms/step - loss: 2.6322e-05 - mae: 0.0018 - mape: 15.5121\n",
      "Epoch 16/200\n",
      "4/4 [==============================] - 3s 723ms/step - loss: 3.4525e-05 - mae: 0.0027 - mape: 21.9589\n",
      "Epoch 17/200\n",
      "4/4 [==============================] - 3s 714ms/step - loss: 4.4347e-05 - mae: 0.0021 - mape: 23.2864\n",
      "Epoch 18/200\n",
      "4/4 [==============================] - 3s 755ms/step - loss: 3.3494e-05 - mae: 0.0023 - mape: 16.8693\n",
      "Epoch 19/200\n",
      "4/4 [==============================] - 3s 790ms/step - loss: 3.1968e-05 - mae: 0.0025 - mape: 29.2191\n",
      "Epoch 20/200\n",
      "4/4 [==============================] - 4s 822ms/step - loss: 4.2614e-05 - mae: 0.0022 - mape: 26.9799\n",
      "Epoch 21/200\n",
      "4/4 [==============================] - 3s 721ms/step - loss: 3.4945e-05 - mae: 0.0024 - mape: 20.1692\n",
      "Epoch 22/200\n",
      "4/4 [==============================] - 3s 727ms/step - loss: 2.7484e-05 - mae: 0.0016 - mape: 12.7620\n",
      "Epoch 23/200\n",
      "4/4 [==============================] - 4s 797ms/step - loss: 3.1063e-05 - mae: 0.0016 - mape: 10.4964\n",
      "Epoch 24/200\n",
      "4/4 [==============================] - 4s 830ms/step - loss: 3.0251e-05 - mae: 0.0022 - mape: 20.2219\n",
      "Epoch 25/200\n",
      "4/4 [==============================] - 4s 840ms/step - loss: 2.9128e-05 - mae: 0.0021 - mape: 22.1547\n",
      "Epoch 26/200\n",
      "4/4 [==============================] - 3s 773ms/step - loss: 4.9200e-05 - mae: 0.0024 - mape: 20.7377\n",
      "Epoch 27/200\n",
      "4/4 [==============================] - 3s 735ms/step - loss: 4.3552e-05 - mae: 0.0024 - mape: 22.2591\n",
      "Epoch 28/200\n",
      "4/4 [==============================] - 3s 766ms/step - loss: 4.1932e-05 - mae: 0.0021 - mape: 22.1218\n",
      "Epoch 29/200\n",
      "4/4 [==============================] - 3s 762ms/step - loss: 2.4966e-05 - mae: 0.0015 - mape: 12.5855\n",
      "Epoch 30/200\n",
      "4/4 [==============================] - 3s 730ms/step - loss: 3.2422e-05 - mae: 0.0021 - mape: 16.3977\n",
      "Epoch 31/200\n",
      "4/4 [==============================] - 3s 688ms/step - loss: 2.7291e-05 - mae: 0.0020 - mape: 12.2298\n",
      "Epoch 32/200\n",
      "4/4 [==============================] - 3s 682ms/step - loss: 3.1734e-05 - mae: 0.0025 - mape: 23.1416\n",
      "Epoch 33/200\n",
      "4/4 [==============================] - 3s 698ms/step - loss: 3.8193e-05 - mae: 0.0016 - mape: 9.7934\n",
      "Epoch 34/200\n",
      "4/4 [==============================] - 3s 691ms/step - loss: 2.6323e-05 - mae: 0.0018 - mape: 10.4529\n",
      "Epoch 35/200\n",
      "4/4 [==============================] - 3s 694ms/step - loss: 3.2303e-05 - mae: 0.0023 - mape: 20.4369\n",
      "Epoch 36/200\n",
      "4/4 [==============================] - 3s 699ms/step - loss: 3.1412e-05 - mae: 0.0018 - mape: 13.5088\n",
      "Epoch 37/200\n",
      "4/4 [==============================] - 3s 740ms/step - loss: 4.4224e-05 - mae: 0.0024 - mape: 15.5813\n",
      "Epoch 38/200\n",
      "4/4 [==============================] - 3s 748ms/step - loss: 5.2860e-05 - mae: 0.0040 - mape: 36.0395\n",
      "Epoch 39/200\n",
      "4/4 [==============================] - 3s 767ms/step - loss: 3.6998e-05 - mae: 0.0030 - mape: 22.5217\n",
      "Epoch 40/200\n",
      "4/4 [==============================] - 3s 742ms/step - loss: 3.0726e-05 - mae: 0.0028 - mape: 22.9313\n",
      "Epoch 41/200\n",
      "4/4 [==============================] - 3s 720ms/step - loss: 6.0390e-05 - mae: 0.0036 - mape: 23.7881\n",
      "Epoch 42/200\n",
      "4/4 [==============================] - 3s 737ms/step - loss: 4.0335e-05 - mae: 0.0039 - mape: 31.6479\n",
      "Epoch 43/200\n",
      "4/4 [==============================] - 3s 804ms/step - loss: 3.5247e-05 - mae: 0.0026 - mape: 30.4368\n",
      "Epoch 44/200\n",
      "4/4 [==============================] - 3s 767ms/step - loss: 3.8661e-05 - mae: 0.0029 - mape: 30.4491\n",
      "Epoch 45/200\n",
      "4/4 [==============================] - 3s 716ms/step - loss: 3.7307e-05 - mae: 0.0025 - mape: 26.3713\n",
      "Epoch 46/200\n",
      "4/4 [==============================] - 3s 731ms/step - loss: 4.5419e-05 - mae: 0.0031 - mape: 36.4504\n",
      "Epoch 47/200\n",
      "4/4 [==============================] - 3s 746ms/step - loss: 3.9292e-05 - mae: 0.0017 - mape: 16.6574\n",
      "Epoch 48/200\n",
      "4/4 [==============================] - 3s 800ms/step - loss: 3.7401e-05 - mae: 0.0018 - mape: 18.1474\n",
      "Epoch 49/200\n",
      "4/4 [==============================] - 3s 768ms/step - loss: 2.7562e-05 - mae: 0.0022 - mape: 16.1907\n",
      "Epoch 50/200\n",
      "4/4 [==============================] - 3s 695ms/step - loss: 3.2454e-05 - mae: 0.0022 - mape: 28.1788\n",
      "Epoch 51/200\n",
      "4/4 [==============================] - 3s 685ms/step - loss: 3.1524e-05 - mae: 0.0020 - mape: 14.4156\n",
      "Epoch 52/200\n",
      "4/4 [==============================] - 3s 675ms/step - loss: 5.0616e-05 - mae: 0.0030 - mape: 33.5059\n",
      "Epoch 53/200\n",
      "4/4 [==============================] - 3s 679ms/step - loss: 3.9890e-05 - mae: 0.0029 - mape: 25.1646\n",
      "Epoch 54/200\n",
      "4/4 [==============================] - 3s 674ms/step - loss: 3.2954e-05 - mae: 0.0034 - mape: 39.6001\n",
      "Epoch 55/200\n",
      "4/4 [==============================] - 3s 676ms/step - loss: 4.0334e-05 - mae: 0.0021 - mape: 23.1517\n",
      "Epoch 56/200\n",
      "4/4 [==============================] - 3s 680ms/step - loss: 3.6071e-05 - mae: 0.0027 - mape: 22.2951\n",
      "Epoch 57/200\n",
      "4/4 [==============================] - 3s 700ms/step - loss: 3.6741e-05 - mae: 0.0018 - mape: 11.6511\n",
      "Epoch 58/200\n",
      "4/4 [==============================] - 3s 771ms/step - loss: 3.7604e-05 - mae: 0.0014 - mape: 10.1545\n",
      "Epoch 59/200\n",
      "4/4 [==============================] - 3s 719ms/step - loss: 2.4605e-05 - mae: 0.0016 - mape: 11.9576\n",
      "Epoch 60/200\n",
      "4/4 [==============================] - 3s 731ms/step - loss: 3.1485e-05 - mae: 0.0021 - mape: 23.8468\n",
      "Epoch 61/200\n",
      "4/4 [==============================] - 3s 712ms/step - loss: 3.3821e-05 - mae: 0.0016 - mape: 9.6359\n",
      "Epoch 62/200\n",
      "4/4 [==============================] - 3s 708ms/step - loss: 2.7306e-05 - mae: 0.0021 - mape: 24.1629\n",
      "Epoch 63/200\n",
      "4/4 [==============================] - 3s 704ms/step - loss: 3.3828e-05 - mae: 0.0023 - mape: 20.3963\n",
      "Epoch 64/200\n",
      "4/4 [==============================] - 3s 706ms/step - loss: 2.7118e-05 - mae: 0.0020 - mape: 20.9060\n",
      "Epoch 65/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.9478e-05 - mae: 0.0022 - mape: 18.9138\n",
      "Epoch 66/200\n",
      "4/4 [==============================] - 3s 701ms/step - loss: 2.5518e-05 - mae: 0.0025 - mape: 24.5762\n",
      "Epoch 67/200\n",
      "4/4 [==============================] - 3s 710ms/step - loss: 3.2144e-05 - mae: 0.0023 - mape: 27.3093\n",
      "Epoch 68/200\n",
      "4/4 [==============================] - 3s 693ms/step - loss: 4.0867e-05 - mae: 0.0027 - mape: 23.5481\n",
      "Epoch 69/200\n",
      "4/4 [==============================] - 3s 699ms/step - loss: 3.8575e-05 - mae: 0.0026 - mape: 21.9898\n",
      "Epoch 70/200\n",
      "4/4 [==============================] - 3s 699ms/step - loss: 4.3849e-05 - mae: 0.0039 - mape: 34.4002\n",
      "Epoch 71/200\n",
      "4/4 [==============================] - 3s 715ms/step - loss: 3.1478e-05 - mae: 0.0021 - mape: 14.4589\n",
      "Epoch 72/200\n",
      "4/4 [==============================] - 3s 708ms/step - loss: 3.0804e-05 - mae: 0.0018 - mape: 13.6006\n",
      "Epoch 73/200\n",
      "4/4 [==============================] - 3s 709ms/step - loss: 2.2010e-05 - mae: 0.0017 - mape: 16.7923\n",
      "Epoch 74/200\n",
      "4/4 [==============================] - 3s 748ms/step - loss: 4.1638e-05 - mae: 0.0026 - mape: 19.0059\n",
      "Epoch 75/200\n",
      "4/4 [==============================] - 3s 730ms/step - loss: 2.8769e-05 - mae: 0.0018 - mape: 12.9388\n",
      "Epoch 76/200\n",
      "4/4 [==============================] - 3s 783ms/step - loss: 2.8448e-05 - mae: 0.0020 - mape: 16.5366\n",
      "Epoch 77/200\n",
      "4/4 [==============================] - 3s 756ms/step - loss: 2.8192e-05 - mae: 0.0023 - mape: 23.9911\n",
      "Epoch 78/200\n",
      "4/4 [==============================] - 3s 716ms/step - loss: 2.7144e-05 - mae: 0.0015 - mape: 9.8122\n",
      "Epoch 79/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.6621e-05 - mae: 0.0016 - mape: 12.8447\n",
      "Epoch 80/200\n",
      "4/4 [==============================] - 3s 807ms/step - loss: 3.7905e-05 - mae: 0.0016 - mape: 11.7887\n",
      "Epoch 81/200\n",
      "4/4 [==============================] - 3s 767ms/step - loss: 2.7565e-05 - mae: 0.0023 - mape: 22.2258\n",
      "Epoch 82/200\n",
      "4/4 [==============================] - 3s 761ms/step - loss: 4.3288e-05 - mae: 0.0029 - mape: 31.5967\n",
      "Epoch 83/200\n",
      "4/4 [==============================] - 3s 746ms/step - loss: 4.6829e-05 - mae: 0.0025 - mape: 17.3516\n",
      "Epoch 84/200\n",
      "4/4 [==============================] - 3s 732ms/step - loss: 2.2386e-05 - mae: 0.0022 - mape: 22.1926\n",
      "Epoch 85/200\n",
      "4/4 [==============================] - 3s 728ms/step - loss: 2.1695e-05 - mae: 0.0014 - mape: 11.5783\n",
      "Epoch 86/200\n",
      "4/4 [==============================] - 3s 768ms/step - loss: 2.7137e-05 - mae: 0.0020 - mape: 14.9011\n",
      "Epoch 87/200\n",
      "4/4 [==============================] - 3s 723ms/step - loss: 2.8418e-05 - mae: 0.0018 - mape: 13.6224\n",
      "Epoch 88/200\n",
      "4/4 [==============================] - 3s 700ms/step - loss: 2.3147e-05 - mae: 0.0015 - mape: 8.9464\n",
      "Epoch 89/200\n",
      "4/4 [==============================] - 3s 769ms/step - loss: 2.9632e-05 - mae: 0.0022 - mape: 22.3420\n",
      "Epoch 90/200\n",
      "4/4 [==============================] - 3s 759ms/step - loss: 2.3780e-05 - mae: 0.0017 - mape: 11.5388\n",
      "Epoch 91/200\n",
      "4/4 [==============================] - 3s 699ms/step - loss: 4.0726e-05 - mae: 0.0020 - mape: 16.9703\n",
      "Epoch 92/200\n",
      "4/4 [==============================] - 3s 790ms/step - loss: 4.9852e-05 - mae: 0.0038 - mape: 38.0726\n",
      "Epoch 93/200\n",
      "4/4 [==============================] - 3s 760ms/step - loss: 3.5273e-05 - mae: 0.0030 - mape: 25.6266\n",
      "Epoch 94/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.6440e-05 - mae: 0.0019 - mape: 18.4036\n",
      "Epoch 95/200\n",
      "4/4 [==============================] - 3s 725ms/step - loss: 3.8107e-05 - mae: 0.0024 - mape: 20.8202\n",
      "Epoch 96/200\n",
      "4/4 [==============================] - 3s 705ms/step - loss: 4.0723e-05 - mae: 0.0021 - mape: 18.8166\n",
      "Epoch 97/200\n",
      "4/4 [==============================] - 3s 708ms/step - loss: 5.6471e-05 - mae: 0.0047 - mape: 42.8505\n",
      "Epoch 98/200\n",
      "4/4 [==============================] - 3s 702ms/step - loss: 4.1546e-05 - mae: 0.0029 - mape: 24.2258\n",
      "Epoch 99/200\n",
      "4/4 [==============================] - 3s 704ms/step - loss: 3.0674e-05 - mae: 0.0022 - mape: 17.9034\n",
      "Epoch 100/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.3694e-05 - mae: 0.0017 - mape: 13.8533\n",
      "Epoch 101/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.2926e-05 - mae: 0.0015 - mape: 15.0249\n",
      "Epoch 102/200\n",
      "4/4 [==============================] - 3s 706ms/step - loss: 2.2107e-05 - mae: 0.0015 - mape: 15.7062\n",
      "Epoch 103/200\n",
      "4/4 [==============================] - 3s 698ms/step - loss: 2.3145e-05 - mae: 0.0016 - mape: 12.3912\n",
      "Epoch 104/200\n",
      "4/4 [==============================] - 3s 701ms/step - loss: 2.8780e-05 - mae: 0.0020 - mape: 20.3726\n",
      "Epoch 105/200\n",
      "4/4 [==============================] - 3s 695ms/step - loss: 3.7909e-05 - mae: 0.0029 - mape: 24.0426\n",
      "Epoch 106/200\n",
      "4/4 [==============================] - 3s 735ms/step - loss: 5.4836e-05 - mae: 0.0042 - mape: 44.4422\n",
      "Epoch 107/200\n",
      "4/4 [==============================] - 3s 727ms/step - loss: 5.0179e-05 - mae: 0.0028 - mape: 26.9143\n",
      "Epoch 108/200\n",
      "4/4 [==============================] - 4s 829ms/step - loss: 3.1685e-05 - mae: 0.0032 - mape: 30.1038\n",
      "Epoch 109/200\n",
      "4/4 [==============================] - 3s 796ms/step - loss: 3.0575e-05 - mae: 0.0034 - mape: 44.1636\n",
      "Epoch 110/200\n",
      "4/4 [==============================] - 3s 742ms/step - loss: 3.0202e-05 - mae: 0.0017 - mape: 14.0467\n",
      "Epoch 111/200\n",
      "4/4 [==============================] - 3s 709ms/step - loss: 3.5743e-05 - mae: 0.0020 - mape: 20.0239\n",
      "Epoch 112/200\n",
      "4/4 [==============================] - 3s 757ms/step - loss: 2.8496e-05 - mae: 0.0017 - mape: 19.1280\n",
      "Epoch 113/200\n",
      "4/4 [==============================] - 3s 802ms/step - loss: 2.3927e-05 - mae: 0.0024 - mape: 22.0042\n",
      "Epoch 114/200\n",
      "4/4 [==============================] - 3s 784ms/step - loss: 3.3200e-05 - mae: 0.0029 - mape: 36.7808\n",
      "Epoch 115/200\n",
      "4/4 [==============================] - 3s 745ms/step - loss: 3.6336e-05 - mae: 0.0030 - mape: 32.7611\n",
      "Epoch 116/200\n",
      "4/4 [==============================] - 3s 733ms/step - loss: 2.5953e-05 - mae: 0.0016 - mape: 15.5367\n",
      "Epoch 117/200\n",
      "4/4 [==============================] - 3s 798ms/step - loss: 4.2547e-05 - mae: 0.0039 - mape: 38.0681\n",
      "Epoch 118/200\n",
      "4/4 [==============================] - 3s 722ms/step - loss: 4.3409e-05 - mae: 0.0037 - mape: 31.4638\n",
      "Epoch 119/200\n",
      "4/4 [==============================] - 3s 695ms/step - loss: 2.8525e-05 - mae: 0.0022 - mape: 17.8085\n",
      "Epoch 120/200\n",
      "4/4 [==============================] - 3s 731ms/step - loss: 3.1323e-05 - mae: 0.0025 - mape: 20.0654\n",
      "Epoch 121/200\n",
      "4/4 [==============================] - 3s 716ms/step - loss: 2.5698e-05 - mae: 0.0016 - mape: 11.0042\n",
      "Epoch 122/200\n",
      "4/4 [==============================] - 3s 701ms/step - loss: 2.2318e-05 - mae: 0.0014 - mape: 10.3158\n",
      "Epoch 123/200\n",
      "4/4 [==============================] - 3s 710ms/step - loss: 2.3846e-05 - mae: 0.0019 - mape: 17.6495\n",
      "Epoch 124/200\n",
      "4/4 [==============================] - 3s 750ms/step - loss: 3.5110e-05 - mae: 0.0019 - mape: 23.3840\n",
      "Epoch 125/200\n",
      "4/4 [==============================] - 3s 694ms/step - loss: 3.4661e-05 - mae: 0.0015 - mape: 9.9602\n",
      "Epoch 126/200\n",
      "4/4 [==============================] - 3s 698ms/step - loss: 1.8391e-05 - mae: 0.0014 - mape: 9.9889\n",
      "Epoch 127/200\n",
      "4/4 [==============================] - 3s 702ms/step - loss: 4.1053e-05 - mae: 0.0022 - mape: 22.9466\n",
      "Epoch 128/200\n",
      "4/4 [==============================] - 3s 695ms/step - loss: 4.7937e-05 - mae: 0.0036 - mape: 38.8236\n",
      "Epoch 129/200\n",
      "4/4 [==============================] - 3s 699ms/step - loss: 2.9901e-05 - mae: 0.0025 - mape: 22.7563\n",
      "Epoch 130/200\n",
      "4/4 [==============================] - 3s 678ms/step - loss: 4.0103e-05 - mae: 0.0019 - mape: 14.7206\n",
      "Epoch 131/200\n",
      "4/4 [==============================] - 3s 681ms/step - loss: 4.5154e-05 - mae: 0.0027 - mape: 21.4970\n",
      "Epoch 132/200\n",
      "4/4 [==============================] - 3s 735ms/step - loss: 3.0389e-05 - mae: 0.0020 - mape: 18.9422\n",
      "Epoch 133/200\n",
      "4/4 [==============================] - 3s 716ms/step - loss: 3.0207e-05 - mae: 0.0022 - mape: 15.9766\n",
      "Epoch 134/200\n",
      "4/4 [==============================] - 3s 747ms/step - loss: 4.0054e-05 - mae: 0.0028 - mape: 29.1458\n",
      "Epoch 135/200\n",
      "4/4 [==============================] - 3s 801ms/step - loss: 3.0128e-05 - mae: 0.0024 - mape: 21.7367\n",
      "Epoch 136/200\n",
      "4/4 [==============================] - 3s 809ms/step - loss: 4.2138e-05 - mae: 0.0041 - mape: 45.1568\n",
      "Epoch 137/200\n",
      "4/4 [==============================] - 4s 814ms/step - loss: 4.4802e-05 - mae: 0.0041 - mape: 47.2749\n",
      "Epoch 138/200\n",
      "4/4 [==============================] - 3s 738ms/step - loss: 4.7107e-05 - mae: 0.0031 - mape: 30.7848\n",
      "Epoch 139/200\n",
      "4/4 [==============================] - 3s 689ms/step - loss: 3.8976e-05 - mae: 0.0028 - mape: 41.9811\n",
      "Epoch 140/200\n",
      "4/4 [==============================] - 3s 657ms/step - loss: 4.3952e-05 - mae: 0.0019 - mape: 14.6406\n",
      "Epoch 141/200\n",
      "4/4 [==============================] - 3s 657ms/step - loss: 2.3485e-05 - mae: 0.0025 - mape: 30.6511\n",
      "Epoch 142/200\n",
      "4/4 [==============================] - 3s 668ms/step - loss: 3.7279e-05 - mae: 0.0033 - mape: 37.4802\n",
      "Epoch 143/200\n",
      "4/4 [==============================] - 3s 660ms/step - loss: 3.1648e-05 - mae: 0.0035 - mape: 33.2857\n",
      "Epoch 144/200\n",
      "4/4 [==============================] - 3s 677ms/step - loss: 5.2110e-05 - mae: 0.0024 - mape: 25.9857\n",
      "Epoch 145/200\n",
      "4/4 [==============================] - 3s 723ms/step - loss: 4.1881e-05 - mae: 0.0028 - mape: 22.3381\n",
      "Epoch 146/200\n",
      "4/4 [==============================] - 3s 820ms/step - loss: 4.0695e-05 - mae: 0.0029 - mape: 28.9777\n",
      "Epoch 147/200\n",
      "4/4 [==============================] - 4s 854ms/step - loss: 4.2233e-05 - mae: 0.0020 - mape: 23.3408\n",
      "Epoch 148/200\n",
      "4/4 [==============================] - 4s 826ms/step - loss: 2.8847e-05 - mae: 0.0020 - mape: 22.0226\n",
      "Epoch 149/200\n",
      "4/4 [==============================] - 3s 726ms/step - loss: 2.4017e-05 - mae: 0.0018 - mape: 19.3387\n",
      "Epoch 150/200\n",
      "4/4 [==============================] - 3s 717ms/step - loss: 2.7360e-05 - mae: 0.0013 - mape: 8.2891\n",
      "Epoch 151/200\n",
      "4/4 [==============================] - 3s 689ms/step - loss: 2.0702e-05 - mae: 0.0015 - mape: 11.8304\n",
      "Epoch 152/200\n",
      "4/4 [==============================] - 3s 715ms/step - loss: 2.9055e-05 - mae: 0.0020 - mape: 18.6520\n",
      "Epoch 153/200\n",
      "4/4 [==============================] - 3s 714ms/step - loss: 2.7280e-05 - mae: 0.0016 - mape: 12.1554\n",
      "Epoch 154/200\n",
      "4/4 [==============================] - 3s 695ms/step - loss: 2.5927e-05 - mae: 0.0015 - mape: 12.4249\n",
      "Epoch 155/200\n",
      "4/4 [==============================] - 3s 708ms/step - loss: 3.6196e-05 - mae: 0.0020 - mape: 22.1538\n",
      "Epoch 156/200\n",
      "4/4 [==============================] - 3s 712ms/step - loss: 2.9507e-05 - mae: 0.0021 - mape: 19.2173\n",
      "Epoch 157/200\n",
      "4/4 [==============================] - 3s 715ms/step - loss: 2.4638e-05 - mae: 0.0014 - mape: 9.6222\n",
      "Epoch 158/200\n",
      "4/4 [==============================] - 3s 718ms/step - loss: 2.4504e-05 - mae: 0.0022 - mape: 19.1775\n",
      "Epoch 159/200\n",
      "4/4 [==============================] - 3s 741ms/step - loss: 4.1633e-05 - mae: 0.0028 - mape: 21.0159\n",
      "Epoch 160/200\n",
      "4/4 [==============================] - 3s 688ms/step - loss: 2.1235e-05 - mae: 0.0022 - mape: 17.1900\n",
      "Epoch 161/200\n",
      "4/4 [==============================] - 3s 688ms/step - loss: 3.8242e-05 - mae: 0.0028 - mape: 18.0455\n",
      "Epoch 162/200\n",
      "4/4 [==============================] - 3s 689ms/step - loss: 4.7949e-05 - mae: 0.0044 - mape: 38.6664\n",
      "Epoch 163/200\n",
      "4/4 [==============================] - 3s 674ms/step - loss: 4.4584e-05 - mae: 0.0032 - mape: 24.7914\n",
      "Epoch 164/200\n",
      "4/4 [==============================] - 3s 690ms/step - loss: 2.4084e-05 - mae: 0.0024 - mape: 29.0747\n",
      "Epoch 165/200\n",
      "4/4 [==============================] - 3s 682ms/step - loss: 3.1353e-05 - mae: 0.0019 - mape: 24.4828\n",
      "Epoch 166/200\n",
      "4/4 [==============================] - 3s 694ms/step - loss: 2.9693e-05 - mae: 0.0025 - mape: 26.5383\n",
      "Epoch 167/200\n",
      "4/4 [==============================] - 3s 700ms/step - loss: 2.3798e-05 - mae: 0.0025 - mape: 28.5134\n",
      "Epoch 168/200\n",
      "4/4 [==============================] - 3s 696ms/step - loss: 2.4885e-05 - mae: 0.0022 - mape: 18.8008\n",
      "Epoch 169/200\n",
      "4/4 [==============================] - 3s 717ms/step - loss: 2.8238e-05 - mae: 0.0022 - mape: 18.0094\n",
      "Epoch 170/200\n",
      "4/4 [==============================] - 3s 708ms/step - loss: 1.7741e-05 - mae: 0.0015 - mape: 9.3619\n",
      "Epoch 171/200\n",
      "4/4 [==============================] - 3s 825ms/step - loss: 5.4859e-05 - mae: 0.0034 - mape: 28.9609\n",
      "Epoch 172/200\n",
      "4/4 [==============================] - 3s 729ms/step - loss: 5.5311e-05 - mae: 0.0058 - mape: 57.0369\n",
      "Epoch 173/200\n",
      "4/4 [==============================] - 3s 806ms/step - loss: 3.1589e-05 - mae: 0.0028 - mape: 21.9843\n",
      "Epoch 174/200\n",
      "4/4 [==============================] - 3s 813ms/step - loss: 2.1553e-05 - mae: 0.0023 - mape: 37.0751\n",
      "Epoch 175/200\n",
      "4/4 [==============================] - 4s 827ms/step - loss: 3.4848e-05 - mae: 0.0017 - mape: 24.8883\n",
      "Epoch 176/200\n",
      "4/4 [==============================] - 3s 755ms/step - loss: 2.8577e-05 - mae: 0.0017 - mape: 15.1855\n",
      "Epoch 177/200\n",
      "4/4 [==============================] - 3s 711ms/step - loss: 3.6934e-05 - mae: 0.0023 - mape: 21.0352\n",
      "Epoch 178/200\n",
      "4/4 [==============================] - 3s 698ms/step - loss: 3.6284e-05 - mae: 0.0021 - mape: 22.0539\n",
      "Epoch 179/200\n",
      "4/4 [==============================] - 3s 799ms/step - loss: 2.8878e-05 - mae: 0.0014 - mape: 8.9149\n",
      "Epoch 180/200\n",
      "4/4 [==============================] - 3s 781ms/step - loss: 2.1165e-05 - mae: 0.0016 - mape: 14.2244\n",
      "Epoch 181/200\n",
      "4/4 [==============================] - 3s 715ms/step - loss: 2.5091e-05 - mae: 0.0022 - mape: 21.5607\n",
      "Epoch 182/200\n",
      "4/4 [==============================] - 3s 727ms/step - loss: 4.6473e-05 - mae: 0.0037 - mape: 34.4309\n",
      "Epoch 183/200\n",
      "4/4 [==============================] - 3s 767ms/step - loss: 2.3590e-05 - mae: 0.0022 - mape: 13.8343\n",
      "Epoch 184/200\n",
      "4/4 [==============================] - 4s 857ms/step - loss: 3.3374e-05 - mae: 0.0021 - mape: 17.7003\n",
      "Epoch 185/200\n",
      "4/4 [==============================] - 3s 760ms/step - loss: 3.0120e-05 - mae: 0.0030 - mape: 24.6085\n",
      "Epoch 186/200\n",
      "4/4 [==============================] - 3s 729ms/step - loss: 2.6749e-05 - mae: 0.0018 - mape: 11.3116\n",
      "Epoch 187/200\n",
      "4/4 [==============================] - 3s 730ms/step - loss: 2.5586e-05 - mae: 0.0014 - mape: 11.7910\n",
      "Epoch 188/200\n",
      "4/4 [==============================] - 3s 725ms/step - loss: 2.6916e-05 - mae: 0.0016 - mape: 11.9157\n",
      "Epoch 189/200\n",
      "4/4 [==============================] - 3s 701ms/step - loss: 2.4646e-05 - mae: 0.0017 - mape: 20.1361\n",
      "Epoch 190/200\n",
      "4/4 [==============================] - 3s 678ms/step - loss: 2.2965e-05 - mae: 0.0015 - mape: 16.5892\n",
      "Epoch 191/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 3.5801e-05 - mae: 0.0020 - mape: 16.9193\n",
      "Epoch 192/200\n",
      "4/4 [==============================] - 3s 718ms/step - loss: 1.9623e-05 - mae: 0.0014 - mape: 8.3648\n",
      "Epoch 193/200\n",
      "4/4 [==============================] - 3s 709ms/step - loss: 3.4739e-05 - mae: 0.0015 - mape: 13.4878\n",
      "Epoch 194/200\n",
      "4/4 [==============================] - 3s 730ms/step - loss: 2.1765e-05 - mae: 0.0016 - mape: 11.8513\n",
      "Epoch 195/200\n",
      "4/4 [==============================] - 3s 723ms/step - loss: 2.6837e-05 - mae: 0.0015 - mape: 8.2184\n",
      "Epoch 196/200\n",
      "4/4 [==============================] - 3s 706ms/step - loss: 2.6082e-05 - mae: 0.0018 - mape: 14.7284\n",
      "Epoch 197/200\n",
      "4/4 [==============================] - 3s 736ms/step - loss: 2.7076e-05 - mae: 0.0020 - mape: 18.0117\n",
      "Epoch 198/200\n",
      "4/4 [==============================] - 3s 703ms/step - loss: 2.3247e-05 - mae: 0.0017 - mape: 15.8385\n",
      "Epoch 199/200\n",
      "4/4 [==============================] - 3s 691ms/step - loss: 2.2110e-05 - mae: 0.0016 - mape: 14.4622\n",
      "Epoch 200/200\n",
      "4/4 [==============================] - 3s 706ms/step - loss: 2.6601e-05 - mae: 0.0023 - mape: 20.4744\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x7fb9dc43d130>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = create_model()\n",
    "model.fit(X_train, y_train, batch_size=2048, epochs=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2/2 [==============================] - 2s 197ms/step - loss: 2.3680e-05 - mae: 0.0025 - mape: 43.4198\n"
     ]
    }
   ],
   "source": [
    "results = model.evaluate(X_test, y_test, batch_size=2048)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.028003701358788297, 0.02693834315492135, 0.02738274971996299, 0.02734926703355574, 0.026197158233088203, 0.025314432864169876, 0.02507701017873667, 0.025632518385038723, 0.02576492719037647, 0.026369137485998156] [[0.02768084]] 0.026758754200555207\n"
     ]
    }
   ],
   "source": [
    "results = model.predict(X_test[1:2])\n",
    "\n",
    "print(X_test[1], results, y_test[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            [(None, 10, 1)]      0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional (Bidirectional)   (None, 10, 256)      133120      input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional_1 (Bidirectional) (None, 10, 256)      394240      bidirectional[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "bidirectional_2 (Bidirectional) (None, 10, 128)      164352      bidirectional_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling1d (Globa (None, 128)          0           bidirectional_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "global_max_pooling1d (GlobalMax (None, 128)          0           bidirectional_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "concatenate (Concatenate)       (None, 256)          0           global_average_pooling1d[0][0]   \n",
      "                                                                 global_max_pooling1d[0][0]       \n",
      "__________________________________________________________________________________________________\n",
      "dense (Dense)                   (None, 64)           16448       concatenate[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 1)            65          dense[0][0]                      \n",
      "==================================================================================================\n",
      "Total params: 708,225\n",
      "Trainable params: 708,225\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "c449169ff25f4796428b4530909a07e6d2d6090a7030f40470c61baf40d583f3"
  },
  "kernelspec": {
   "display_name": "Python 3.8.5 ('base')",
   "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": "3.8.5"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}