01_ucb1_modeling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "from striatum.bandit import ucb1\n",
    "from utils.data import *\n",
    "from utils.policy import *\n",
    "plt.style.use('seaborn')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The dataset can be downloaded from http://bit.ly/2Q1kr0T."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_DIR = os.path.abspath('../../Data/yahoo!_front_page_today_module_user_click_log/R6')\n",
    "DATA_FILE_NAMES = ['ydata-fp-td-clicks-v1_0.2009050{}.gz'.format(str(d + 3)) for d in range(7)]\n",
    "MODEL_DIR = os.path.abspath('models')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_paths = []\n",
    "for data_file in DATA_FILE_NAMES:\n",
    "    data_paths.append(os.path.join(DATA_DIR, data_file))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "policy_learner = PolicyLearner(ucb1.UCB1, seed=0, reset_freq=1000000)\n",
    "param_grid = {'data_size': [0.01, 0.05, 0.1, 0.2, 0.3, 1.0]}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: {'data_size': 0.01} | Average Reward: 3.8559% | Computation Time: 2:41:25\n",
      "Parameters: {'data_size': 0.05} | Average Reward: 4.2268% | Computation Time: 2:44:36\n",
      "Parameters: {'data_size': 0.1} | Average Reward: 4.1464% | Computation Time: 3:28:02\n",
      "Parameters: {'data_size': 0.2} | Average Reward: 4.2165% | Computation Time: 3:29:33\n",
      "Parameters: {'data_size': 0.3} | Average Reward: 4.3314% | Computation Time: 4:11:33\n",
      "Parameters: {'data_size': 1.0} | Average Reward: 4.5450% | Computation Time: 5:44:50\n"
     ]
    }
   ],
   "source": [
    "res = search_param(policy_learner, param_grid, parse_data, data_paths=data_paths)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.isdir(MODEL_DIR):\n",
    "    os.mkdir(MODEL_DIR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(os.path.join(MODEL_DIR, 'model_ucb1.pkl'), 'wb') as file:\n",
    "    pickle.dump(res, file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(os.path.join(MODEL_DIR, 'model_ucb1.pkl'), 'rb') as file:\n",
    "    res = pickle.load(file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameter: {'data_size': 1}\n",
      "Cumulative CTR: 4.5450%\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "best_param = res['param'][np.argmax(res['avg_reward'])]\n",
    "cum_ctr = np.max(res['avg_reward'])\n",
    "print('Best Parameter: {0}\\nCumulative CTR: {1:0.4%}'.format(best_param, cum_ctr))\n",
    "fig, ax = plt.subplots(figsize=(10, 6))\n",
    "_ = ax.plot(res['seq_avg_reward'][np.argmax(res['avg_reward'])], color='mediumblue', label='UCB1', linewidth=1)\n",
    "ax.axhline(y=cum_ctr, color='black', linestyle='--', linewidth=1)\n",
    "ax.set_xlabel('Num. of Events')\n",
    "ax.set_ylabel('CTR')\n",
    "ax.set_ylim([0.0, 0.1])\n",
    "ax.legend(loc='best')\n",
    "_ = ax.set_title('Average Reward', fontsize=12, fontweight='bold')"
   ]
  }
 ],
 "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": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}