disan_ts_pred.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from disan_keras_layers import DISAN\n",
    "from disan_keras import get_attn, plot_attn\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from keras.models import Model\n",
    "from keras.layers import Input, Dense, Reshape\n",
    "from keras.optimizers import Adadelta\n",
    "from keras.callbacks import ReduceLROnPlateau, ModelCheckpoint, TensorBoard, EarlyStopping\n",
    "import pandas_datareader.data as web\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "import os\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Open</th>\n",
       "      <th>High</th>\n",
       "      <th>Low</th>\n",
       "      <th>Close</th>\n",
       "      <th>Adj Close</th>\n",
       "      <th>Volume</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Date</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2010-01-04</th>\n",
       "      <td>112.370003</td>\n",
       "      <td>113.389999</td>\n",
       "      <td>111.510002</td>\n",
       "      <td>113.330002</td>\n",
       "      <td>96.341499</td>\n",
       "      <td>118944600</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-01-05</th>\n",
       "      <td>113.260002</td>\n",
       "      <td>113.680000</td>\n",
       "      <td>112.849998</td>\n",
       "      <td>113.629997</td>\n",
       "      <td>96.596519</td>\n",
       "      <td>111579900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-01-06</th>\n",
       "      <td>113.519997</td>\n",
       "      <td>113.989998</td>\n",
       "      <td>113.430000</td>\n",
       "      <td>113.709999</td>\n",
       "      <td>96.664566</td>\n",
       "      <td>116074400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-01-07</th>\n",
       "      <td>113.500000</td>\n",
       "      <td>114.330002</td>\n",
       "      <td>113.180000</td>\n",
       "      <td>114.190002</td>\n",
       "      <td>97.072586</td>\n",
       "      <td>131091100</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2010-01-08</th>\n",
       "      <td>113.889999</td>\n",
       "      <td>114.620003</td>\n",
       "      <td>113.660004</td>\n",
       "      <td>114.570000</td>\n",
       "      <td>97.395630</td>\n",
       "      <td>126402800</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  Open        High         Low       Close  Adj Close  \\\n",
       "Date                                                                    \n",
       "2010-01-04  112.370003  113.389999  111.510002  113.330002  96.341499   \n",
       "2010-01-05  113.260002  113.680000  112.849998  113.629997  96.596519   \n",
       "2010-01-06  113.519997  113.989998  113.430000  113.709999  96.664566   \n",
       "2010-01-07  113.500000  114.330002  113.180000  114.190002  97.072586   \n",
       "2010-01-08  113.889999  114.620003  113.660004  114.570000  97.395630   \n",
       "\n",
       "               Volume  \n",
       "Date                   \n",
       "2010-01-04  118944600  \n",
       "2010-01-05  111579900  \n",
       "2010-01-06  116074400  \n",
       "2010-01-07  131091100  \n",
       "2010-01-08  126402800  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = web.get_data_yahoo('SPY')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEACAYAAAC9Gb03AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x15da5940>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ts = df[['Adj Close', 'Open', 'High', 'Low']].dropna()\n",
    "ts.plot(subplots=True);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-0.09847840609974945 0.05095638777688283 (2058, 4)\n",
      "(1843, 10, 4) (1843,) (205, 10, 4) (205,)\n"
     ]
    }
   ],
   "source": [
    "t = 10\n",
    "train_perc = 0.9\n",
    "\n",
    "#ts_scaled = MinMaxScaler().fit_transform(ts.values.reshape(-1, 1))[:, 0]\n",
    "ts_scaled = np.diff(np.log(ts.values + 1), axis=0)\n",
    "print(ts_scaled.min(), ts_scaled.max(), ts_scaled.shape)\n",
    "\n",
    "X, y = [], []\n",
    "for i in range(len(ts_scaled)):\n",
    "    if i >= t:\n",
    "        X.append(ts_scaled[i-t:i])\n",
    "        y.append(ts_scaled[i, 0])\n",
    "        \n",
    "X, y = np.asarray(X), np.asarray(y)\n",
    "train_len = int(len(X) * train_perc)\n",
    "X_train, y_train = X[:train_len, ...], y[:train_len, ...]\n",
    "X_test, y_test = X[train_len:, ...], y[train_len:, ...]\n",
    "\n",
    "print(X_train.shape, y_train.shape, X_test.shape, y_test.shape)\n",
    "assert not np.isnan(ts_scaled).any() and not np.isnan(ts_scaled).any() and not np.isinf(ts_scaled).any() and not np.isinf(ts_scaled).any()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_1 (InputLayer)         (None, 10, 4)             0         \n",
      "_________________________________________________________________\n",
      "disan_1 (DISAN)              (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 1)                 129       \n",
      "=================================================================\n",
      "Total params: 129\n",
      "Trainable params: 129\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "n = 64\n",
    "\n",
    "ts_input = Input(shape=(t, 4))\n",
    "#ts_reshape = Reshape((t, 1))(ts_input)\n",
    "\n",
    "disan = DISAN(n, dropout=0.2)\n",
    "rep_2d = disan(ts_input)\n",
    "\n",
    "output = Dense(1, activation='linear')(rep_2d)\n",
    "\n",
    "model = Model(ts_input, output)\n",
    "opt = Adadelta(lr=0.5)\n",
    "model.compile(loss='mse', optimizer=opt, metrics=['mae'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1658 samples, validate on 185 samples\n",
      "Epoch 1/100\n",
      "1658/1658 [==============================] - 3s 2ms/step - loss: 1.0151e-04 - mean_absolute_error: 0.0071 - val_loss: 2.9743e-05 - val_mean_absolute_error: 0.0041\n",
      "Epoch 2/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0170e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8740e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00002: ReduceLROnPlateau reducing learning rate to 0.05.\n",
      "Epoch 3/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0175e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8653e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00003: ReduceLROnPlateau reducing learning rate to 0.005000000074505806.\n",
      "Epoch 4/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0119e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00004: ReduceLROnPlateau reducing learning rate to 0.0004999999888241291.\n",
      "Epoch 5/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0206e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00005: ReduceLROnPlateau reducing learning rate to 4.9999996554106475e-05.\n",
      "Epoch 6/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0130e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00006: ReduceLROnPlateau reducing learning rate to 4.999999509891496e-06.\n",
      "Epoch 7/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0118e-04 - mean_absolute_error: 0.0070 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00007: ReduceLROnPlateau reducing learning rate to 4.999999418942025e-07.\n",
      "Epoch 8/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0166e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00008: ReduceLROnPlateau reducing learning rate to 4.999999418942025e-08.\n",
      "Epoch 9/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0138e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00009: ReduceLROnPlateau reducing learning rate to 4.999999347887752e-09.\n",
      "Epoch 10/100\n",
      "1658/1658 [==============================] - 2s 1ms/step - loss: 1.0130e-04 - mean_absolute_error: 0.0071 - val_loss: 2.8652e-05 - val_mean_absolute_error: 0.0039\n",
      "\n",
      "Epoch 00010: ReduceLROnPlateau reducing learning rate to 4.999999525523436e-10.\n",
      "Epoch 00010: early stopping\n",
      "205/205 [==============================] - 0s 1ms/step\n",
      "[4.850892582908273e-05, 0.004240086767822504]\n"
     ]
    }
   ],
   "source": [
    "filepath = 'models'\n",
    "callbacks = [\n",
    "    ReduceLROnPlateau(patience=0, verbose=1),\n",
    "    EarlyStopping(patience=3, verbose=1),\n",
    "    #ModelCheckpoint(filepath, verbose=1, save_best_only=True, save_weights_only=True),\n",
    "    #TensorBoard(os.path.join(filepath, 'logs'), histogram_freq=1)\n",
    "]\n",
    "\n",
    "model.fit(X_train, y_train, validation_split=0.1, epochs=100, verbose=1, callbacks=callbacks)\n",
    "print(model.evaluate(X_test, y_test, batch_size=len(X_test)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "forward_attn_full (10, 10, 64)\n",
      "forward_attn_fusion (10, 64)\n",
      "backward_attn_full (10, 10, 64)\n",
      "backward_attn_fusion (10, 64)\n",
      "multidim_attn (10, 128)\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAe8AAAFbCAYAAAD80gauAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1801e780>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "attn = get_attn(X_test[np.newaxis, np.random.randint(len(X_test)), ...], model, disan.attn_dict)\n",
    "for k, v in attn.items():\n",
    "    print(k, v.shape)\n",
    "\n",
    "plot_attn(attn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(205,)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred = model.predict(X_test)[:, 0]\n",
    "y_pred.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAD8CAYAAABkbJM/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nB4aZmWVxYJiZWRYHhpmZZXFgmJlZlv8P/6FSLjG8FoUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x17f7dac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(y_test, y_pred);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": "3.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}