stock-prediction.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Stock Prediction with Recurrent Neural Network\n",
    "\n",
    "Deep learning is involved a lot in the modern quantitive financial field. There are many different neural networks can be applied to stock price prediction problems. The recurrent neural network, to be specific, the Long Short Term Memory(LSTM) network outperforms others architecture since it can take advantage of predicting time series (or sequentially) involved result with a specific configuration.\n",
    "\n",
    "We will make a really simple LSTM with Keras to predict the stock price in the Chinese stock."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import time\n",
    "import math\n",
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Dropout, Activation\n",
    "from keras.layers.recurrent import LSTM\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import sklearn.preprocessing as prep"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import Data\n",
    "\n",
    "I have introduced how we get this dataset both in README.md and in another notebook."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>volume</th>\n",
       "      <th>amount</th>\n",
       "      <th>close</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>20.64</td>\n",
       "      <td>20.64</td>\n",
       "      <td>20.37</td>\n",
       "      <td>16362363.0</td>\n",
       "      <td>3.350027e+08</td>\n",
       "      <td>20.56</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>20.92</td>\n",
       "      <td>20.92</td>\n",
       "      <td>20.60</td>\n",
       "      <td>21850597.0</td>\n",
       "      <td>4.520071e+08</td>\n",
       "      <td>20.64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>21.00</td>\n",
       "      <td>21.15</td>\n",
       "      <td>20.72</td>\n",
       "      <td>26910139.0</td>\n",
       "      <td>5.628396e+08</td>\n",
       "      <td>20.94</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>20.70</td>\n",
       "      <td>21.57</td>\n",
       "      <td>20.70</td>\n",
       "      <td>64585536.0</td>\n",
       "      <td>1.363421e+09</td>\n",
       "      <td>21.02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>20.60</td>\n",
       "      <td>20.70</td>\n",
       "      <td>20.20</td>\n",
       "      <td>45886018.0</td>\n",
       "      <td>9.382043e+08</td>\n",
       "      <td>20.70</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    open   high    low      volume        amount  close\n",
       "0  20.64  20.64  20.37  16362363.0  3.350027e+08  20.56\n",
       "1  20.92  20.92  20.60  21850597.0  4.520071e+08  20.64\n",
       "2  21.00  21.15  20.72  26910139.0  5.628396e+08  20.94\n",
       "3  20.70  21.57  20.70  64585536.0  1.363421e+09  21.02\n",
       "4  20.60  20.70  20.20  45886018.0  9.382043e+08  20.70"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv('000002-from-1995-01-01.csv')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preprocess Data\n",
    "\n",
    "We can see from the table above that the scales of our features vary a lot. It will be beneficial to normalize your training data before you feed data into your model. \"Having different features with widely different scales fed to your model will cause the network to weight the features not equally. This can cause a falsely prioritisation of some features over the others in the representation.\" Read [here](http://stackoverflow.com/a/43526354/6670143) for more detailed explanation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "def standard_scaler(X_train, X_test):\n",
    "    train_samples, train_nx, train_ny = X_train.shape\n",
    "    test_samples, test_nx, test_ny = X_test.shape\n",
    "    \n",
    "    X_train = X_train.reshape((train_samples, train_nx * train_ny))\n",
    "    X_test = X_test.reshape((test_samples, test_nx * test_ny))\n",
    "    \n",
    "    preprocessor = prep.StandardScaler().fit(X_train)\n",
    "    X_train = preprocessor.transform(X_train)\n",
    "    X_test = preprocessor.transform(X_test)\n",
    "    \n",
    "    X_train = X_train.reshape((train_samples, train_nx, train_ny))\n",
    "    X_test = X_test.reshape((test_samples, test_nx, test_ny))\n",
    "    \n",
    "    return X_train, X_test"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Split the data to X_train, y_train, X_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preprocess_data(stock, seq_len):\n",
    "    amount_of_features = len(stock.columns)\n",
    "    data = stock.values\n",
    "    \n",
    "    sequence_length = seq_len + 1\n",
    "    result = []\n",
    "    for index in range(len(data) - sequence_length):\n",
    "        result.append(data[index : index + sequence_length])\n",
    "        \n",
    "    result = np.array(result)\n",
    "    row = round(0.9 * result.shape[0])\n",
    "    train = result[: int(row), :]\n",
    "    \n",
    "    train, result = standard_scaler(train, result)\n",
    "    \n",
    "    X_train = train[:, : -1]\n",
    "    y_train = train[:, -1][: ,-1]\n",
    "    X_test = result[int(row) :, : -1]\n",
    "    y_test = result[int(row) :, -1][ : ,-1]\n",
    "\n",
    "    X_train = np.reshape(X_train, (X_train.shape[0], X_train.shape[1], amount_of_features))\n",
    "    X_test = np.reshape(X_test, (X_test.shape[0], X_test.shape[1], amount_of_features))  \n",
    "\n",
    "    return [X_train, y_train, X_test, y_test]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build the LSTM Network\n",
    "\n",
    "Here we will build a simple RNN with 2 LSTM layers.\n",
    "The architecture is:\n",
    "    \n",
    "    LSTM --> Dropout --> LSTM --> Dropout --> Fully-Conneted(Dense)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_model(layers):\n",
    "    model = Sequential()\n",
    "\n",
    "    # By setting return_sequences to True we are able to stack another LSTM layer\n",
    "    model.add(LSTM(\n",
    "        input_dim=layers[0],\n",
    "        output_dim=layers[1],\n",
    "        return_sequences=True))\n",
    "    model.add(Dropout(0.4))\n",
    "\n",
    "    model.add(LSTM(\n",
    "        layers[2],\n",
    "        return_sequences=False))\n",
    "    model.add(Dropout(0.3))\n",
    "\n",
    "    model.add(Dense(\n",
    "        output_dim=layers[3]))\n",
    "    model.add(Activation(\"linear\"))\n",
    "\n",
    "    start = time.time()\n",
    "    model.compile(loss=\"mse\", optimizer=\"rmsprop\", metrics=['accuracy'])\n",
    "    print(\"Compilation Time : \", time.time() - start)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train (4648, 20, 6)\n",
      "y_train (4648,)\n",
      "X_test (516, 20, 6)\n",
      "y_test (516,)\n"
     ]
    }
   ],
   "source": [
    "window = 20\n",
    "X_train, y_train, X_test, y_test = preprocess_data(df[:: -1], window)\n",
    "print(\"X_train\", X_train.shape)\n",
    "print(\"y_train\", y_train.shape)\n",
    "print(\"X_test\", X_test.shape)\n",
    "print(\"y_test\", y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/weston/.pyenv/versions/3.6.1/envs/nn/lib/python3.6/site-packages/ipykernel_launcher.py:8: UserWarning: The `input_dim` and `input_length` arguments in recurrent layers are deprecated. Use `input_shape` instead.\n",
      "  \n",
      "/Users/weston/.pyenv/versions/3.6.1/envs/nn/lib/python3.6/site-packages/ipykernel_launcher.py:8: UserWarning: Update your `LSTM` call to the Keras 2 API: `LSTM(return_sequences=True, input_shape=(None, 6), units=20)`\n",
      "  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Compilation Time :  0.02150106430053711\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/weston/.pyenv/versions/3.6.1/envs/nn/lib/python3.6/site-packages/ipykernel_launcher.py:17: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(units=1)`\n"
     ]
    }
   ],
   "source": [
    "model = build_model([X_train.shape[2], window, 100, 1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Training the Network\n",
    "\n",
    "You can play with the hyperparameters here. The training process can take a significant time if you are using a average GPU or CPU(worse!)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x11afb5668>"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(\n",
    "    X_train,\n",
    "    y_train,\n",
    "    batch_size=768,\n",
    "    epochs=5,\n",
    "    validation_split=0.1,\n",
    "    verbose=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Score: 0.04 MSE (0.19 RMSE)\n",
      "Test Score: 0.11 MSE (0.33 RMSE)\n"
     ]
    }
   ],
   "source": [
    "trainScore = model.evaluate(X_train, y_train, verbose=0)\n",
    "print('Train Score: %.2f MSE (%.2f RMSE)' % (trainScore[0], math.sqrt(trainScore[0])))\n",
    "\n",
    "testScore = model.evaluate(X_test, y_test, verbose=0)\n",
    "print('Test Score: %.2f MSE (%.2f RMSE)' % (testScore[0], math.sqrt(testScore[0])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize the Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "diff = []\n",
    "ratio = []\n",
    "pred = model.predict(X_test)\n",
    "for u in range(len(y_test)):\n",
    "    pr = pred[u][0]\n",
    "    ratio.append((y_test[u] / pr) - 1)\n",
    "    diff.append(abs(y_test[u] - pr))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Scale it back"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# if using within a jupyter notebook\n",
    "%matplotlib inline \n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt2.plot(pred, color='red', label='Prediction')\n",
    "plt2.plot(y_test, color='blue', label='Ground Truth')\n",
    "plt2.legend(loc='upper left')\n",
    "plt2.show()"
   ]
  }
 ],
 "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": "3.6.1"
  },
  "widgets": {
   "state": {},
   "version": "1.1.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}