model_LSTM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Here I use LSTM model for multivariate time-series prediction, the model output is day by day."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from pandas import DataFrame, Series\n",
    "import matplotlib.pyplot as plt\n",
    "from numpy import concatenate\n",
    "\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.metrics import mean_squared_error"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM\n",
    "from keras.layers import Dense"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cite Jason ML's function to process time-series data for supervised learning\n",
    "https://machinelearningmastery.com/multivariate-time-series-forecasting-lstms-keras/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def series_to_supervised(data, n_in=1, n_out=1, dropnan=True):\n",
    "\t\"\"\"\n",
    "\tFrame a time series as a supervised learning dataset.\n",
    "\tArguments:\n",
    "\t\tdata: Sequence of observations as a list or NumPy array.\n",
    "\t\tn_in: Number of lag observations as input (X).\n",
    "\t\tn_out: Number of observations as output (y).\n",
    "\t\tdropnan: Boolean whether or not to drop rows with NaN values.\n",
    "\tReturns:\n",
    "\t\tPandas DataFrame of series framed for supervised learning.\n",
    "\t\"\"\"\n",
    "\tn_vars = 1 if type(data) is list else data.shape[1]\n",
    "\tdf = DataFrame(data)\n",
    "\tcols, names = list(), list()\n",
    "\t# input sequence (t-n, ... t-1)\n",
    "\tfor i in range(n_in, 0, -1):\n",
    "\t\tcols.append(df.shift(i))\n",
    "\t\tnames += [('var%d(t-%d)' % (j+1, i)) for j in range(n_vars)]\n",
    "\t# forecast sequence (t, t+1, ... t+n)\n",
    "\tfor i in range(0, n_out):\n",
    "\t\tcols.append(df.shift(-i))\n",
    "\t\tif i == 0:\n",
    "\t\t\tnames += [('var%d(t)' % (j+1)) for j in range(n_vars)]\n",
    "\t\telse:\n",
    "\t\t\tnames += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)]\n",
    "\t# put it all together\n",
    "\tagg = concat(cols, axis=1)\n",
    "\tagg.columns = names\n",
    "\t# drop rows with NaN values\n",
    "\tif dropnan:\n",
    "\t\tagg.dropna(inplace=True)\n",
    "\treturn agg"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load data and exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x3a9e208>"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0xe7ee400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "soybean = pd.read_csv('ALL_Prices_soybean.csv', index_col = 'Date', parse_dates=True).sort_index()\n",
    "# examine NA\n",
    "soybean.isna().sum(axis=0)\n",
    "# describe soybean\n",
    "soybean.describe()\n",
    "# drop duplicates\n",
    "soybean.drop_duplicates(inplace=True)\n",
    "\n",
    "# Market_Open = 1\n",
    "soybean = soybean.loc[soybean['Market_Open'] == 1]\n",
    "# soybean price plot\n",
    "soybean.plot(y='bean_settle')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature engineering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "rolling_feature = ['bean_settle']\n",
    "rolling_wins = [5, 10, 20, 60]\n",
    "for var in rolling_feature:\n",
    "    for num in rolling_wins:\n",
    "        soybean[var+str(num)] = soybean[var].rolling(num).aggregate(np.mean)\n",
    "\n",
    "keep = ['bean_settle', 'bean_high', 'bean_low', 'bean_settle5', 'bean_settle10', 'bean_settle20', 'bean_settle60',\n",
    "        'Gold', 'USD', 'Oil']    \n",
    "soybean = soybean.filter(items=keep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# normalizing numerical features\n",
    "soybean.dropna(inplace=True)\n",
    "StdScaler = StandardScaler()\n",
    "soybean.loc[:, keep] = StdScaler.fit_transform(soybean.loc[:, keep])\n",
    "soybean.loc[:, keep] = soybean.loc[:, keep].astype('float32')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pre-processing for LSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x182decc0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "reframed = series_to_supervised(soybean.values, 1, 1)\n",
    "\n",
    "#%% Split data\n",
    "reframed.drop(reframed.columns[np.arange(11,20)], axis=1, inplace=True)\n",
    "reframed = reframed.values\n",
    "\n",
    "train, test = reframed[:round(len(reframed)*.95), :], reframed[round(len(reframed)*.95):, :] \n",
    "x_train, y_train = train[:, :-1], train[:, -1]\n",
    "x_test, y_test = test[:, :-1], test[:, -1]\n",
    "\n",
    "plt.plot(x_train[:,0])\n",
    "plt.plot(x_test[:,0])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(13500, 1, 10) (13500,) (711, 1, 10) (711,)\n"
     ]
    }
   ],
   "source": [
    "x_train = x_train.reshape((x_train.shape[0], 1, x_train.shape[1]))\n",
    "x_test = x_test.reshape((x_test.shape[0], 1, x_test.shape[1]))\n",
    "print(x_train.shape, y_train.shape, x_test.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LSTM model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 13500 samples, validate on 711 samples\n",
      "Epoch 1/100\n",
      " - 2s - loss: 0.9961 - val_loss: 1.0567\n",
      "Epoch 2/100\n",
      " - 0s - loss: 0.9821 - val_loss: 1.0477\n",
      "Epoch 3/100\n",
      " - 0s - loss: 0.9688 - val_loss: 1.0360\n",
      "Epoch 4/100\n",
      " - 0s - loss: 0.9554 - val_loss: 1.0210\n",
      "Epoch 5/100\n",
      " - 0s - loss: 0.9411 - val_loss: 1.0046\n",
      "Epoch 6/100\n",
      " - 0s - loss: 0.9264 - val_loss: 0.9880\n",
      "Epoch 7/100\n",
      " - 0s - loss: 0.9105 - val_loss: 0.9699\n",
      "Epoch 8/100\n",
      " - 0s - loss: 0.8922 - val_loss: 0.9486\n",
      "Epoch 9/100\n",
      " - 0s - loss: 0.8713 - val_loss: 0.9246\n",
      "Epoch 10/100\n",
      " - 0s - loss: 0.8478 - val_loss: 0.8977\n",
      "Epoch 11/100\n",
      " - 0s - loss: 0.8207 - val_loss: 0.8677\n",
      "Epoch 12/100\n",
      " - 0s - loss: 0.7895 - val_loss: 0.8342\n",
      "Epoch 13/100\n",
      " - 0s - loss: 0.7533 - val_loss: 0.7965\n",
      "Epoch 14/100\n",
      " - 0s - loss: 0.7101 - val_loss: 0.7543\n",
      "Epoch 15/100\n",
      " - 0s - loss: 0.6597 - val_loss: 0.7070\n",
      "Epoch 16/100\n",
      " - 0s - loss: 0.6018 - val_loss: 0.6542\n",
      "Epoch 17/100\n",
      " - 0s - loss: 0.5365 - val_loss: 0.5956\n",
      "Epoch 18/100\n",
      " - 0s - loss: 0.4654 - val_loss: 0.5319\n",
      "Epoch 19/100\n",
      " - 0s - loss: 0.3924 - val_loss: 0.4646\n",
      "Epoch 20/100\n",
      " - 0s - loss: 0.3238 - val_loss: 0.3972\n",
      "Epoch 21/100\n",
      " - 0s - loss: 0.2708 - val_loss: 0.3358\n",
      "Epoch 22/100\n",
      " - 0s - loss: 0.2377 - val_loss: 0.2898\n",
      "Epoch 23/100\n",
      " - 0s - loss: 0.2122 - val_loss: 0.2676\n",
      "Epoch 24/100\n",
      " - 0s - loss: 0.1796 - val_loss: 0.2675\n",
      "Epoch 25/100\n",
      " - 0s - loss: 0.1432 - val_loss: 0.2787\n",
      "Epoch 26/100\n",
      " - 0s - loss: 0.1109 - val_loss: 0.2902\n",
      "Epoch 27/100\n",
      " - 0s - loss: 0.0858 - val_loss: 0.2958\n",
      "Epoch 28/100\n",
      " - 0s - loss: 0.0693 - val_loss: 0.2942\n",
      "Epoch 29/100\n",
      " - 0s - loss: 0.0624 - val_loss: 0.2870\n",
      "Epoch 30/100\n",
      " - 0s - loss: 0.0623 - val_loss: 0.2770\n",
      "Epoch 31/100\n",
      " - 0s - loss: 0.0628 - val_loss: 0.2669\n",
      "Epoch 32/100\n",
      " - 0s - loss: 0.0609 - val_loss: 0.2584\n",
      "Epoch 33/100\n",
      " - 0s - loss: 0.0583 - val_loss: 0.2519\n",
      "Epoch 34/100\n",
      " - 0s - loss: 0.0564 - val_loss: 0.2469\n",
      "Epoch 35/100\n",
      " - 0s - loss: 0.0552 - val_loss: 0.2424\n",
      "Epoch 36/100\n",
      " - 0s - loss: 0.0540 - val_loss: 0.2377\n",
      "Epoch 37/100\n",
      " - 0s - loss: 0.0528 - val_loss: 0.2323\n",
      "Epoch 38/100\n",
      " - 0s - loss: 0.0515 - val_loss: 0.2263\n",
      "Epoch 39/100\n",
      " - 0s - loss: 0.0503 - val_loss: 0.2198\n",
      "Epoch 40/100\n",
      " - 0s - loss: 0.0491 - val_loss: 0.2132\n",
      "Epoch 41/100\n",
      " - 0s - loss: 0.0480 - val_loss: 0.2066\n",
      "Epoch 42/100\n",
      " - 0s - loss: 0.0468 - val_loss: 0.2002\n",
      "Epoch 43/100\n",
      " - 0s - loss: 0.0457 - val_loss: 0.1940\n",
      "Epoch 44/100\n",
      " - 0s - loss: 0.0445 - val_loss: 0.1877\n",
      "Epoch 45/100\n",
      " - 0s - loss: 0.0434 - val_loss: 0.1814\n",
      "Epoch 46/100\n",
      " - 0s - loss: 0.0423 - val_loss: 0.1749\n",
      "Epoch 47/100\n",
      " - 0s - loss: 0.0411 - val_loss: 0.1684\n",
      "Epoch 48/100\n",
      " - 0s - loss: 0.0400 - val_loss: 0.1617\n",
      "Epoch 49/100\n",
      " - 0s - loss: 0.0389 - val_loss: 0.1550\n",
      "Epoch 50/100\n",
      " - 0s - loss: 0.0378 - val_loss: 0.1483\n",
      "Epoch 51/100\n",
      " - 0s - loss: 0.0367 - val_loss: 0.1416\n",
      "Epoch 52/100\n",
      " - 0s - loss: 0.0356 - val_loss: 0.1350\n",
      "Epoch 53/100\n",
      " - 0s - loss: 0.0345 - val_loss: 0.1285\n",
      "Epoch 54/100\n",
      " - 0s - loss: 0.0335 - val_loss: 0.1222\n",
      "Epoch 55/100\n",
      " - 0s - loss: 0.0324 - val_loss: 0.1159\n",
      "Epoch 56/100\n",
      " - 0s - loss: 0.0314 - val_loss: 0.1096\n",
      "Epoch 57/100\n",
      " - 0s - loss: 0.0304 - val_loss: 0.1035\n",
      "Epoch 58/100\n",
      " - 0s - loss: 0.0294 - val_loss: 0.0976\n",
      "Epoch 59/100\n",
      " - 0s - loss: 0.0284 - val_loss: 0.0918\n",
      "Epoch 60/100\n",
      " - 0s - loss: 0.0275 - val_loss: 0.0862\n",
      "Epoch 61/100\n",
      " - 0s - loss: 0.0266 - val_loss: 0.0808\n",
      "Epoch 62/100\n",
      " - 0s - loss: 0.0257 - val_loss: 0.0755\n",
      "Epoch 63/100\n",
      " - 0s - loss: 0.0248 - val_loss: 0.0704\n",
      "Epoch 64/100\n",
      " - 0s - loss: 0.0239 - val_loss: 0.0656\n",
      "Epoch 65/100\n",
      " - 0s - loss: 0.0230 - val_loss: 0.0609\n",
      "Epoch 66/100\n",
      " - 0s - loss: 0.0222 - val_loss: 0.0565\n",
      "Epoch 67/100\n",
      " - 0s - loss: 0.0214 - val_loss: 0.0523\n",
      "Epoch 68/100\n",
      " - 0s - loss: 0.0206 - val_loss: 0.0483\n",
      "Epoch 69/100\n",
      " - 0s - loss: 0.0198 - val_loss: 0.0445\n",
      "Epoch 70/100\n",
      " - 0s - loss: 0.0191 - val_loss: 0.0409\n",
      "Epoch 71/100\n",
      " - 0s - loss: 0.0183 - val_loss: 0.0375\n",
      "Epoch 72/100\n",
      " - 0s - loss: 0.0176 - val_loss: 0.0343\n",
      "Epoch 73/100\n",
      " - 0s - loss: 0.0169 - val_loss: 0.0314\n",
      "Epoch 74/100\n",
      " - 0s - loss: 0.0162 - val_loss: 0.0287\n",
      "Epoch 75/100\n",
      " - 0s - loss: 0.0156 - val_loss: 0.0261\n",
      "Epoch 76/100\n",
      " - 0s - loss: 0.0149 - val_loss: 0.0237\n",
      "Epoch 77/100\n",
      " - 0s - loss: 0.0143 - val_loss: 0.0215\n",
      "Epoch 78/100\n",
      " - 0s - loss: 0.0137 - val_loss: 0.0194\n",
      "Epoch 79/100\n",
      " - 0s - loss: 0.0131 - val_loss: 0.0175\n",
      "Epoch 80/100\n",
      " - 0s - loss: 0.0126 - val_loss: 0.0158\n",
      "Epoch 81/100\n",
      " - 0s - loss: 0.0121 - val_loss: 0.0142\n",
      "Epoch 82/100\n",
      " - 0s - loss: 0.0116 - val_loss: 0.0129\n",
      "Epoch 83/100\n",
      " - 0s - loss: 0.0111 - val_loss: 0.0116\n",
      "Epoch 84/100\n",
      " - 0s - loss: 0.0107 - val_loss: 0.0105\n",
      "Epoch 85/100\n",
      " - 0s - loss: 0.0102 - val_loss: 0.0095\n",
      "Epoch 86/100\n",
      " - 0s - loss: 0.0098 - val_loss: 0.0087\n",
      "Epoch 87/100\n",
      " - 0s - loss: 0.0094 - val_loss: 0.0079\n",
      "Epoch 88/100\n",
      " - 0s - loss: 0.0091 - val_loss: 0.0072\n",
      "Epoch 89/100\n",
      " - 0s - loss: 0.0088 - val_loss: 0.0067\n",
      "Epoch 90/100\n",
      " - 0s - loss: 0.0084 - val_loss: 0.0062\n",
      "Epoch 91/100\n",
      " - 0s - loss: 0.0082 - val_loss: 0.0057\n",
      "Epoch 92/100\n",
      " - 0s - loss: 0.0079 - val_loss: 0.0053\n",
      "Epoch 93/100\n",
      " - 0s - loss: 0.0076 - val_loss: 0.0050\n",
      "Epoch 94/100\n",
      " - 0s - loss: 0.0074 - val_loss: 0.0048\n",
      "Epoch 95/100\n",
      " - 0s - loss: 0.0072 - val_loss: 0.0045\n",
      "Epoch 96/100\n",
      " - 0s - loss: 0.0070 - val_loss: 0.0043\n",
      "Epoch 97/100\n",
      " - 0s - loss: 0.0068 - val_loss: 0.0042\n",
      "Epoch 98/100\n",
      " - 0s - loss: 0.0066 - val_loss: 0.0040\n",
      "Epoch 99/100\n",
      " - 0s - loss: 0.0065 - val_loss: 0.0039\n",
      "Epoch 100/100\n",
      " - 0s - loss: 0.0063 - val_loss: 0.0038\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "#  1st hidden layer\n",
    "model.add(LSTM(30, input_shape=(x_train.shape[1], x_train.shape[2]), activation='relu', return_sequences=True))\n",
    "#  2nd hidden layer\n",
    "model.add(LSTM(15, activation='relu'))\n",
    "# output layer\n",
    "model.add(Dense(1))\n",
    "# metric & optimizer\n",
    "model.compile(loss='mean_squared_error', optimizer='adam')\n",
    "# fit model\n",
    "history = model.fit(x_train, y_train, epochs=100, batch_size=round(len(x_train)*0.2), validation_data=(x_test, y_test), verbose=2, shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### plot training error"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1845bf98>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1b5d4240>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'], label='train')\n",
    "plt.plot(history.history['val_loss'], label='test')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "\n",
    "plt.plot(np.abs(np.array(history.history['loss']) - np.array(history.history['val_loss'])), label='train/test diff')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# performance evaluation\n",
    "def mean_absolute_percentage_error(y_true, y_pred): \n",
    "    return np.mean(np.abs((y_true - y_pred) / y_true)) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "# prediction\n",
    "yhat = model.predict(x_test)\n",
    "x_test = x_test.reshape((x_test.shape[0], x_test.shape[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test MSE: 350.34 and MAPE: 1.43%\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1b63c048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# inverse transform y_hat\n",
    "inv_yhat = concatenate((yhat, x_test[:, 1:]), axis=1)\n",
    "inv_yhat = StdScaler.inverse_transform(inv_yhat)\n",
    "y_pred = inv_yhat[:,0]\n",
    "\n",
    "# inver transform y\n",
    "y_test = y_test.reshape((len(y_test), 1))\n",
    "inv_y = concatenate((y_test, x_test[:, 1:]), axis=1)\n",
    "inv_y = StdScaler.inverse_transform(inv_y)\n",
    "y_true = inv_y[:,0]\n",
    "\n",
    "mse_test = mean_squared_error(y_true, y_pred)\n",
    "mape_test = mean_absolute_percentage_error(y_true, y_pred)\n",
    "print(\"Test MSE: {:.2f} and MAPE: {:.2f}%\".format(mse_test, mape_test))\n",
    "\n",
    "# plot y_pred vs y_true\n",
    "plt.plot(y_pred, 'r', label='prediction')\n",
    "plt.plot(y_true, 'b', label='true')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Short summary"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Adding one more layer in LSTM model would further improve MAPE."
   ]
  }
 ],
 "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}