FB predictions using LSTM

notebook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from numpy import array\n",
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM\n",
    "from keras.layers import Dense, Dropout, Activation\n",
    "from tensorflow import keras\n",
    "import tensorflow as tf\n",
    "from sklearn import metrics\n",
    "import datetime\n",
    "pd.options.plotting.backend = \"plotly\"\n",
    "%load_ext tensorboard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_parquet(\"./data/raw_historical_data/FB.parquet\")\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split a univariate sequence into samples\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)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# choose a number of time steps\n",
    "n_steps = 50\n",
    "# split into samples\n",
    "X, y = split_sequence(df.adjClose, n_steps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_size = int(0.7 * X.shape[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, y_train, X_validate, y_validate = X[:train_size], y[:train_size], X[train_size + n_steps + 1:], y[train_size + n_steps + 1:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# reshape from [samples, timesteps] into [samples, timesteps, features]\n",
    "n_features = 1\n",
    "X_train = X_train.reshape(X_train.shape[0], X_train.shape[1], n_features)\n",
    "X_validate = X_validate.reshape(X_validate.shape[0], X_validate.shape[1], n_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "log_dir = \"logs/fit/\" + datetime.datetime.now().strftime(\"%Y%m%d-%H%M%S\")\n",
    "tensorboard_callback = tf.keras.callbacks.TensorBoard(log_dir=log_dir, histogram_freq=1)\n",
    "\n",
    "# define model\n",
    "model = Sequential()\n",
    "model.add(LSTM(50, activation=keras.activations.relu, input_shape=(n_steps, n_features)))\n",
    "# model.add(Dropout(0.2, name='lstm_dropout_0'))\n",
    "# model.add(Dense(64, name='dense_0'))\n",
    "# model.add(Activation('sigmoid', name='sigmoid_0'))\n",
    "model.add(Dense(1))\n",
    "model.compile(optimizer=keras.optimizers.Adam(learning_rate=0.01), loss='mse')\n",
    "# fit model\n",
    "model.fit(X_train, y_train, epochs=200, verbose=1, callbacks=[tensorboard_callback])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# demonstrate prediction\n",
    "yhat = model.predict(X_validate, verbose=0)\n",
    "yhat = yhat.reshape(yhat.shape[0])\n",
    "\n",
    "y_vals = pd.DataFrame(\n",
    "    {\n",
    "        \"true\": y_validate,\n",
    "        \"pred\": yhat\n",
    "    }\n",
    ")\n",
    "y_vals.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_vals.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "(\n",
    "    f\"Mean absolute error: {metrics.mean_absolute_error(y_validate, yhat)}  /  \" + \n",
    "    f\"Mean squared error: {metrics.mean_squared_error(y_validate, yhat)}\"\n",
    ")"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "ab3782f7a3367a0ac78615a23ccde97a675b35a20d9404343327068ff2cfd379"
  },
  "kernelspec": {
   "display_name": "Python 3.9.6 64-bit ('stocks-opportunities-monitor': conda)",
   "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": "3.9.6"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}