LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "uuid": "bfe09417-5fdf-427a-a1f9-41b2ea7eb1c9"
   },
   "outputs": [],
   "source": [
    "#Coded by Ferdin Joe John Joseph\n",
    "#DSW Notebook Service in Alibaba Cloud with GPU instance used\n",
    "# vanilla lstm example\n",
    "from numpy import array\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "uuid": "e36e5b80-af16-4b9c-8075-3cb470a515ec"
   },
   "outputs": [],
   "source": [
    "# split a univariate sequence\n",
    "def split_sequence(sequence, n_steps):\n",
    "\tX, y = list(), list()\n",
    "\tfor i in range(len(sequence)):\n",
    "\t\t# find the end of this pattern\n",
    "\t\tend_ix = i + n_steps\n",
    "\t\t# check if we are beyond the sequence\n",
    "\t\tif end_ix > len(sequence)-1:\n",
    "\t\t\tbreak\n",
    "\t\t# gather input and output parts of the pattern\n",
    "\t\tseq_x, seq_y = sequence[i:end_ix], sequence[end_ix]\n",
    "\t\tX.append(seq_x)\n",
    "\t\ty.append(seq_y)\n",
    "\treturn array(X), array(y)\n",
    "\n",
    "# define input sequence\n",
    "raw_seq = [19,14,18,19,20,18,19,28,15,33,49,30,22]\n",
    "#,37,33,24,14,27,12,10,18,13,13,16,13,9,21,22,21,6,8,10,4,7,10,29,4,3,18,23,17,3,9,8,17,9,7,11,1,0,5,5,7,6,0,2,2,6.12,2.3,1.69,1.39,0.79,0.83,0.77,0.68,0.62,0.53,0.44,0.47,0.36,0.33,0.3,0.27,0.32,0.24,0.19,0.25,0.2,0.16,0.13,0.1,0.26,0.19,0.15,0.16,0.17,0.14,0.1,0.13,0.15,0.15,0.12,0.12,0.14,0.12]\n",
    "# choose a number of time steps\n",
    "n_steps = 10\n",
    "# split into samples\n",
    "X, y = split_sequence(raw_seq, n_steps)\n",
    "# reshape from [samples, timesteps] into [samples, timesteps, features]\n",
    "n_features = 1\n",
    "X = X.reshape((X.shape[0], X.shape[1], n_features))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "uuid": "d65117f8-df62-46ea-aa3d-4683d1c99873"
   },
   "outputs": [],
   "source": [
    "# define model\n",
    "model = keras.Sequential()\n",
    "\n",
    "model.add(tf.keras.layers.LSTM(50, activation='relu', return_sequences=True, input_shape=(n_steps, n_features)))\n",
    "model.add(tf.keras.layers.LSTM(50, activation='relu'))\n",
    "model.add(tf.keras.layers.Dense(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "uuid": "5a79a84b-d856-4925-8c7d-493c448f635e"
   },
   "outputs": [],
   "source": [
    "#Compile model\n",
    "model.compile(optimizer='adam', loss='mse')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "uuid": "3c024987-c158-4e88-a35e-ad4eff585cac"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_8\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_3 (LSTM)                (None, 10, 50)            10400     \n",
      "_________________________________________________________________\n",
      "lstm_4 (LSTM)                (None, 50)                20200     \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 51        \n",
      "=================================================================\n",
      "Total params: 30,651\n",
      "Trainable params: 30,651\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "uuid": "c4c3ead2-af82-4ba2-b980-c5572a7099d8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[43.077778]]\n",
      "[[43.142365]]\n",
      "[[27.899595]]\n",
      "[[24.340622]]\n",
      "[[46.4218]]\n",
      "[[12.520472]]\n",
      "[[10.317375]]\n",
      "[[21.999908]]\n",
      "[[12.767969]]\n",
      "[[11.124883]]\n",
      "[[66.73734]]\n",
      "[[28.234184]]\n",
      "[[8.597376]]\n",
      "[[20.508944]]\n",
      "[[25.914412]]\n",
      "[[30.025087]]\n",
      "[[9.279014]]\n",
      "[[22.568993]]\n",
      "[[24.693583]]\n",
      "[[11.5277605]]\n",
      "[[8.85549]]\n",
      "[[8.796094]]\n",
      "[[24.883224]]\n",
      "[[34.26453]]\n",
      "[[3.956057]]\n",
      "[[7.769548]]\n",
      "[[16.823374]]\n",
      "[[11.465925]]\n",
      "[[10.079477]]\n",
      "[[11.955182]]\n",
      "[[14.400595]]\n",
      "[[11.356669]]\n",
      "[[9.536088]]\n",
      "[[7.901217]]\n",
      "[[28.369463]]\n",
      "[[17.343502]]\n",
      "[[12.769805]]\n",
      "[[7.1211433]]\n",
      "[[9.610364]]\n",
      "[[11.4864435]]\n",
      "[[2.4152596]]\n",
      "[[5.486658]]\n",
      "[[7.4775815]]\n",
      "[[2.347696]]\n",
      "[[6.6945934]]\n",
      "[[-0.7078321]]\n",
      "[[1.6982937]]\n",
      "[[2.0744581]]\n",
      "[[1.2150124]]\n",
      "[[1.2347069]]\n",
      "[[0.36732966]]\n",
      "[[0.02947373]]\n",
      "[[0.7231707]]\n",
      "[[0.4463992]]\n",
      "[[0.0427584]]\n",
      "[[0.6861452]]\n",
      "[[0.4238125]]\n",
      "[[0.29608113]]\n",
      "[[0.2881115]]\n",
      "[[0.26376525]]\n",
      "[[0.26686382]]\n",
      "[[0.2518981]]\n",
      "[[0.22226298]]\n",
      "[[0.2524955]]\n",
      "[[0.20535566]]\n",
      "[[0.20163864]]\n",
      "[[0.18501401]]\n",
      "[[0.17453]]\n",
      "[[0.16947463]]\n",
      "[[0.17534691]]\n",
      "[[0.16190587]]\n",
      "[[0.13971028]]\n",
      "[[0.15971944]]\n",
      "[[0.14570057]]\n",
      "[[0.1404157]]\n",
      "[[0.13278076]]\n",
      "[[0.12532195]]\n",
      "[[0.14876762]]\n",
      "[[0.13089794]]\n",
      "[[0.13764049]]\n",
      "[[0.12989765]]\n",
      "[[0.13639976]]\n",
      "[array([[43.077778]], dtype=float32), array([[43.142365]], dtype=float32), array([[27.899595]], dtype=float32), array([[24.340622]], dtype=float32), array([[46.4218]], dtype=float32), array([[12.520472]], dtype=float32), array([[10.317375]], dtype=float32), array([[21.999908]], dtype=float32), array([[12.767969]], dtype=float32), array([[11.124883]], dtype=float32), array([[66.73734]], dtype=float32), array([[28.234184]], dtype=float32), array([[8.597376]], dtype=float32), array([[20.508944]], dtype=float32), array([[25.914412]], dtype=float32), array([[30.025087]], dtype=float32), array([[9.279014]], dtype=float32), array([[22.568993]], dtype=float32), array([[24.693583]], dtype=float32), array([[11.5277605]], dtype=float32), array([[8.85549]], dtype=float32), array([[8.796094]], dtype=float32), array([[24.883224]], dtype=float32), array([[34.26453]], dtype=float32), array([[3.956057]], dtype=float32), array([[7.769548]], dtype=float32), array([[16.823374]], dtype=float32), array([[11.465925]], dtype=float32), array([[10.079477]], dtype=float32), array([[11.955182]], dtype=float32), array([[14.400595]], dtype=float32), array([[11.356669]], dtype=float32), array([[9.536088]], dtype=float32), array([[7.901217]], dtype=float32), array([[28.369463]], dtype=float32), array([[17.343502]], dtype=float32), array([[12.769805]], dtype=float32), array([[7.1211433]], dtype=float32), array([[9.610364]], dtype=float32), array([[11.4864435]], dtype=float32), array([[2.4152596]], dtype=float32), array([[5.486658]], dtype=float32), array([[7.4775815]], dtype=float32), array([[2.347696]], dtype=float32), array([[6.6945934]], dtype=float32), array([[-0.7078321]], dtype=float32), array([[1.6982937]], dtype=float32), array([[2.0744581]], dtype=float32), array([[1.2150124]], dtype=float32), array([[1.2347069]], dtype=float32), array([[0.36732966]], dtype=float32), array([[0.02947373]], dtype=float32), array([[0.7231707]], dtype=float32), array([[0.4463992]], dtype=float32), array([[0.0427584]], dtype=float32), array([[0.6861452]], dtype=float32), array([[0.4238125]], dtype=float32), array([[0.29608113]], dtype=float32), array([[0.2881115]], dtype=float32), array([[0.26376525]], dtype=float32), array([[0.26686382]], dtype=float32), array([[0.2518981]], dtype=float32), array([[0.22226298]], dtype=float32), array([[0.2524955]], dtype=float32), array([[0.20535566]], dtype=float32), array([[0.20163864]], dtype=float32), array([[0.18501401]], dtype=float32), array([[0.17453]], dtype=float32), array([[0.16947463]], dtype=float32), array([[0.17534691]], dtype=float32), array([[0.16190587]], dtype=float32), array([[0.13971028]], dtype=float32), array([[0.15971944]], dtype=float32), array([[0.14570057]], dtype=float32), array([[0.1404157]], dtype=float32), array([[0.13278076]], dtype=float32), array([[0.12532195]], dtype=float32), array([[0.14876762]], dtype=float32), array([[0.13089794]], dtype=float32), array([[0.13764049]], dtype=float32), array([[0.12989765]], dtype=float32), array([[0.13639976]], dtype=float32), 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n"
     ]
    }
   ],
   "source": [
    "# split a univariate sequence\n",
    "def split_sequence(sequence, n_steps):\n",
    "\tX, y = list(), list()\n",
    "\tfor i in range(len(sequence)):\n",
    "\t\t# find the end of this pattern\n",
    "\t\tend_ix = i + n_steps\n",
    "\t\t# check if we are beyond the sequence\n",
    "\t\tif end_ix > len(sequence)-1:\n",
    "\t\t\tbreak\n",
    "\t\t# gather input and output parts of the pattern\n",
    "\t\tseq_x, seq_y = sequence[i:end_ix], sequence[end_ix]\n",
    "\t\tX.append(seq_x)\n",
    "\t\ty.append(seq_y)\n",
    "\treturn array(X), array(y)\n",
    "\n",
    "# define input sequence\n",
    "buckets1 = []\n",
    "for i in range(0,100):\n",
    "    buckets1.append(0)\n",
    "raw_seq = [19,14,18,19,20,18,19,28,15,33,49]\n",
    "fut=[37,33,24,14,27,12,10,18,13,13,16,13,9,21,22,21,6,8,10,4,7,10,29,4,3,18,23,17,3,9,8,17,9,7,11,1,0,5,5,7,6,0,2,2,6.12,2.3,1.69,1.39,0.79,0.83,0.77,0.68,0.62,0.53,0.44,0.47,0.36,0.33,0.3,0.27,0.32,0.24,0.19,0.25,0.2,0.16,0.13,0.1,0.26,0.19,0.15,0.16,0.17,0.14,0.1,0.13,0.15,0.15,0.12,0.12,0.14,0.12]\n",
    "for i in range(len(fut)):\n",
    "\t# choose a number of time steps\n",
    "\tn_steps = 10\n",
    "\t# split into samples\n",
    "\traw_seq.append(fut[i])\n",
    "\tX, y = split_sequence(raw_seq, n_steps)\n",
    "\t# reshape from [samples, timesteps] into [samples, timesteps, features]\n",
    "\tn_features = 1\n",
    "\tX = X.reshape((X.shape[0], X.shape[1], n_features))\n",
    "\t# fit model\n",
    "\tmodel.fit(X, y, epochs=200, verbose=0)\n",
    "\t# demonstrate prediction\n",
    "\tN=10\n",
    "\tx_input = array(raw_seq[-N:])\n",
    "\tx_input = x_input.reshape((1, n_steps, n_features))\n",
    "\tyhat = model.predict(x_input, verbose=0)\n",
    "\tbuckets1[i]=yhat\n",
    "\tprint(yhat)\n",
    "print(buckets1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "uuid": "5510677a-cfe2-47d6-979d-aa03dd89479f"
   },
   "outputs": [],
   "source": [
    "#Stacked LSTM\n",
    "model = keras.Sequential()\n",
    "model.add(tf.keras.layers.LSTM(50, activation='relu', input_shape=(n_steps, n_features)))\n",
    "model.add(tf.keras.layers.Dense(1))\n",
    "model.compile(optimizer='adam', loss='mse')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "uuid": "2e4ce49c-c2c0-460b-b606-ddf93c2c9403"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "lstm_2 (LSTM)                (None, 50)                10400     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 51        \n",
      "=================================================================\n",
      "Total params: 10,451\n",
      "Trainable params: 10,451\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "uuid": "2e83a86c-93fd-4696-9824-a4b5cfebc76d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[47.363373]]\n",
      "[[50.841026]]\n",
      "[[27.357998]]\n",
      "[[20.856781]]\n",
      "[[82.099846]]\n",
      "[[23.026745]]\n",
      "[[21.212326]]\n",
      "[[29.746925]]\n",
      "[[11.468154]]\n",
      "[[14.702788]]\n",
      "[[15.753585]]\n",
      "[[23.317648]]\n",
      "[[9.648972]]\n",
      "[[10.612822]]\n",
      "[[23.493322]]\n",
      "[[27.41619]]\n",
      "[[17.670637]]\n",
      "[[18.465761]]\n",
      "[[12.900002]]\n",
      "[[2.4310033]]\n",
      "[[8.975431]]\n",
      "[[11.275642]]\n",
      "[[41.22436]]\n",
      "[[-2.5949116]]\n",
      "[[4.8990445]]\n",
      "[[0.54645497]]\n",
      "[[16.243689]]\n",
      "[[6.108824]]\n",
      "[[6.973977]]\n",
      "[[14.044634]]\n",
      "[[12.380156]]\n",
      "[[4.5001025]]\n",
      "[[10.707065]]\n",
      "[[8.346088]]\n",
      "[[20.598818]]\n",
      "[[7.6926603]]\n",
      "[[7.1461163]]\n",
      "[[14.755668]]\n",
      "[[3.7570255]]\n",
      "[[10.722033]]\n",
      "[[16.40934]]\n",
      "[[4.960985]]\n",
      "[[0.5576769]]\n",
      "[[5.125187]]\n",
      "[[2.0423386]]\n",
      "[[0.6045375]]\n",
      "[[1.7207422]]\n",
      "[[1.5543057]]\n",
      "[[5.595835]]\n",
      "[[-0.15618856]]\n",
      "[[1.4683285]]\n",
      "[[0.29764628]]\n",
      "[[0.35263973]]\n",
      "[[0.27038562]]\n",
      "[[0.37746274]]\n",
      "[[0.4972602]]\n",
      "[[0.32797566]]\n",
      "[[0.25431505]]\n",
      "[[0.28682524]]\n",
      "[[0.27100402]]\n",
      "[[0.2930026]]\n",
      "[[0.26555985]]\n",
      "[[0.22657838]]\n",
      "[[0.3309465]]\n",
      "[[0.2455784]]\n",
      "[[0.21826889]]\n",
      "[[0.1878809]]\n",
      "[[0.12974481]]\n",
      "[[0.15939157]]\n",
      "[[0.15803006]]\n",
      "[[0.1498235]]\n",
      "[[0.14775392]]\n",
      "[[0.15696599]]\n",
      "[[0.14895089]]\n",
      "[[0.14138871]]\n",
      "[[0.13792706]]\n",
      "[[0.14909132]]\n",
      "[[0.14678147]]\n",
      "[[0.13624507]]\n",
      "[[0.12772536]]\n",
      "[[0.12878191]]\n",
      "[[0.1280053]]\n",
      "[array([[47.363373]], dtype=float32), array([[50.841026]], dtype=float32), array([[27.357998]], dtype=float32), array([[20.856781]], dtype=float32), array([[82.099846]], dtype=float32), array([[23.026745]], dtype=float32), array([[21.212326]], dtype=float32), array([[29.746925]], dtype=float32), array([[11.468154]], dtype=float32), array([[14.702788]], dtype=float32), array([[15.753585]], dtype=float32), array([[23.317648]], dtype=float32), array([[9.648972]], dtype=float32), array([[10.612822]], dtype=float32), array([[23.493322]], dtype=float32), array([[27.41619]], dtype=float32), array([[17.670637]], dtype=float32), array([[18.465761]], dtype=float32), array([[12.900002]], dtype=float32), array([[2.4310033]], dtype=float32), array([[8.975431]], dtype=float32), array([[11.275642]], dtype=float32), array([[41.22436]], dtype=float32), array([[-2.5949116]], dtype=float32), array([[4.8990445]], dtype=float32), array([[0.54645497]], dtype=float32), array([[16.243689]], dtype=float32), array([[6.108824]], dtype=float32), array([[6.973977]], dtype=float32), array([[14.044634]], dtype=float32), array([[12.380156]], dtype=float32), array([[4.5001025]], dtype=float32), array([[10.707065]], dtype=float32), array([[8.346088]], dtype=float32), array([[20.598818]], dtype=float32), array([[7.6926603]], dtype=float32), array([[7.1461163]], dtype=float32), array([[14.755668]], dtype=float32), array([[3.7570255]], dtype=float32), array([[10.722033]], dtype=float32), array([[16.40934]], dtype=float32), array([[4.960985]], dtype=float32), array([[0.5576769]], dtype=float32), array([[5.125187]], dtype=float32), array([[2.0423386]], dtype=float32), array([[0.6045375]], dtype=float32), array([[1.7207422]], dtype=float32), array([[1.5543057]], dtype=float32), array([[5.595835]], dtype=float32), array([[-0.15618856]], dtype=float32), array([[1.4683285]], dtype=float32), array([[0.29764628]], dtype=float32), array([[0.35263973]], dtype=float32), array([[0.27038562]], dtype=float32), array([[0.37746274]], dtype=float32), array([[0.4972602]], dtype=float32), array([[0.32797566]], dtype=float32), array([[0.25431505]], dtype=float32), array([[0.28682524]], dtype=float32), array([[0.27100402]], dtype=float32), array([[0.2930026]], dtype=float32), array([[0.26555985]], dtype=float32), array([[0.22657838]], dtype=float32), array([[0.3309465]], dtype=float32), array([[0.2455784]], dtype=float32), array([[0.21826889]], dtype=float32), array([[0.1878809]], dtype=float32), array([[0.12974481]], dtype=float32), array([[0.15939157]], dtype=float32), array([[0.15803006]], dtype=float32), array([[0.1498235]], dtype=float32), array([[0.14775392]], dtype=float32), array([[0.15696599]], dtype=float32), array([[0.14895089]], dtype=float32), array([[0.14138871]], dtype=float32), array([[0.13792706]], dtype=float32), array([[0.14909132]], dtype=float32), array([[0.14678147]], dtype=float32), array([[0.13624507]], dtype=float32), array([[0.12772536]], dtype=float32), array([[0.12878191]], dtype=float32), array([[0.1280053]], dtype=float32), 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]\n"
     ]
    }
   ],
   "source": [
    "# split a univariate sequence\n",
    "def split_sequence(sequence, n_steps):\n",
    "\tX, y = list(), list()\n",
    "\tfor i in range(len(sequence)):\n",
    "\t\t# find the end of this pattern\n",
    "\t\tend_ix = i + n_steps\n",
    "\t\t# check if we are beyond the sequence\n",
    "\t\tif end_ix > len(sequence)-1:\n",
    "\t\t\tbreak\n",
    "\t\t# gather input and output parts of the pattern\n",
    "\t\tseq_x, seq_y = sequence[i:end_ix], sequence[end_ix]\n",
    "\t\tX.append(seq_x)\n",
    "\t\ty.append(seq_y)\n",
    "\treturn array(X), array(y)\n",
    "\n",
    "# define input sequence\n",
    "buckets = []\n",
    "for i in range(0,100):\n",
    "    buckets.append(0)\n",
    "raw_seq = [19,14,18,19,20,18,19,28,15,33,49]\n",
    "fut=[37,33,24,14,27,12,10,18,13,13,16,13,9,21,22,21,6,8,10,4,7,10,29,4,3,18,23,17,3,9,8,17,9,7,11,1,0,5,5,7,6,0,2,2,6.12,2.3,1.69,1.39,0.79,0.83,0.77,0.68,0.62,0.53,0.44,0.47,0.36,0.33,0.3,0.27,0.32,0.24,0.19,0.25,0.2,0.16,0.13,0.1,0.26,0.19,0.15,0.16,0.17,0.14,0.1,0.13,0.15,0.15,0.12,0.12,0.14,0.12]\n",
    "for i in range(len(fut)):\n",
    "\t# choose a number of time steps\n",
    "\tn_steps = 10\n",
    "\t# split into samples\n",
    "\traw_seq.append(fut[i])\n",
    "\tX, y = split_sequence(raw_seq, n_steps)\n",
    "\t# reshape from [samples, timesteps] into [samples, timesteps, features]\n",
    "\tn_features = 1\n",
    "\tX = X.reshape((X.shape[0], X.shape[1], n_features))\n",
    "\t# fit model\n",
    "\tmodel.fit(X, y, epochs=200, verbose=0)\n",
    "\t# demonstrate prediction\n",
    "\tN=10\n",
    "\tx_input = array(raw_seq[-N:])\n",
    "\tx_input = x_input.reshape((1, n_steps, n_features))\n",
    "\tyhat = model.predict(x_input, verbose=0)\n",
    "\tbuckets[i]=yhat\n",
    "\tprint(yhat)\n",
    "print(buckets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "uuid": "91ed27f5-54e4-4e09-b69d-f0176e8a179f"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "actual=fut\n",
    "vanilla=buckets1\n",
    "stackedl=buckets\n",
    "plt.plot(actual,label='Actual')\n",
    "plt.plot(vanilla,label='Vanilla LSTM forecast')\n",
    "plt.plot(stackedl,label='Stacked LSTM forecast')\n",
    "plt.legend()\n",
    "#display plot\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "uuid": "5a73daa8-9398-4c17-8260-bab8328970ba"
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}