5. Machine Learning with Over-sampling.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings, gc\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import auc, roc_curve\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "# h2o modules\n",
    "import h2o\n",
    "from h2o.frame import H2OFrame\n",
    "from h2o.grid.grid_search import H2OGridSearch\n",
    "from h2o.estimators.glm import H2OGeneralizedLinearEstimator\n",
    "from h2o.estimators.random_forest import H2ORandomForestEstimator\n",
    "\n",
    "# LightGBM modules\n",
    "import lightgbm as lgb\n",
    "\n",
    "# CatBoost modules\n",
    "from catboost import Pool, cv, CatBoostClassifier\n",
    "\n",
    "warnings.simplefilter('ignore')\n",
    "pd.options.display.max_rows = 100\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train:\t (722142, 67)\n",
      "Test:\t (98727, 67) \n",
      "\n",
      "Target 0:\t 549951 \t 0.7616\n",
      "Target 1:\t 172191 \t 0.2384\n",
      "0/1 Ratio:\t 3.1938\n"
     ]
    }
   ],
   "source": [
    "train = pd.read_csv('./data/train_clean.csv')\n",
    "test = pd.read_csv('./data/test_clean.csv')\n",
    "\n",
    "# imbalanced dataset\n",
    "target1 = train['target'].sum()\n",
    "target0 = (1 - train['target']).sum()\n",
    "\n",
    "print('Train:\\t', train.shape)\n",
    "print('Test:\\t', test.shape, '\\n')\n",
    "\n",
    "print('Target 0:\\t', target0, '\\t', np.round(target0 / len(train), 4))\n",
    "print('Target 1:\\t', target1, '\\t', np.round(target1 / len(train), 4))\n",
    "print('0/1 Ratio:\\t', np.round(target0 / target1, 4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Categorical feature:\t 8\n",
      "Numerical feature:\t 56\n",
      "Total feature:\t\t 64\n"
     ]
    }
   ],
   "source": [
    "# define categorical and numerical features\n",
    "cat_features = ['term', 'home_ownership', 'verification_status', 'purpose', \n",
    "                'title', 'addr_state', 'initial_list_status', 'application_type']\n",
    "\n",
    "num_features = ['sub_grade', 'loan_amnt', 'loan_to_inc', 'int_rate', 'installment_ratio', \n",
    "                'emp_length', 'annual_inc', 'dti', 'delinq_2yrs', 'inq_last_6mths', \n",
    "                'open_acc', 'pub_rec', 'revol_bal', 'revol_util', 'total_acc', \n",
    "                'collections_12_mths_ex_med', 'acc_now_delinq', 'tot_coll_amt', \n",
    "                'tot_cur_bal', 'total_rev_hi_lim', 'acc_open_past_24mths', \n",
    "                'avg_cur_bal', 'bc_open_to_buy', 'bc_util', 'chargeoff_within_12_mths',\n",
    "                'delinq_amnt', 'mo_sin_old_il_acct','mo_sin_old_rev_tl_op', \n",
    "                'mo_sin_rcnt_rev_tl_op', 'mo_sin_rcnt_tl', 'mort_acc', \n",
    "                'mths_since_recent_bc', 'mths_since_recent_inq', 'num_accts_ever_120_pd', \n",
    "                'num_actv_bc_tl', 'num_actv_rev_tl', 'num_bc_sats', 'num_bc_tl', \n",
    "                'num_il_tl', 'num_op_rev_tl', 'num_rev_accts', 'num_rev_tl_bal_gt_0', \n",
    "                'num_sats', 'num_tl_120dpd_2m', 'num_tl_30dpd', 'num_tl_90g_dpd_24m', \n",
    "                'num_tl_op_past_12m', 'pct_tl_nvr_dlq', 'percent_bc_gt_75', \n",
    "                'pub_rec_bankruptcies', 'tax_liens', 'tot_hi_cred_lim', 'credit_length',\n",
    "                'total_bal_ex_mort', 'total_bc_limit', 'total_il_high_credit_limit']\n",
    "\n",
    "features = cat_features + num_features\n",
    "\n",
    "# define numerical and categorical features\n",
    "print('Categorical feature:\\t', len(cat_features))\n",
    "print('Numerical feature:\\t', len(num_features))\n",
    "print('Total feature:\\t\\t', len(features))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# I. H2O Data Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Checking whether there is an H2O instance running at http://localhost:54321..... not found.\n",
      "Attempting to start a local H2O server...\n",
      "  Java Version: openjdk version \"1.8.0_171\"; OpenJDK Runtime Environment (build 1.8.0_171-8u171-b11-0ubuntu0.16.04.1-b11); OpenJDK 64-Bit Server VM (build 25.171-b11, mixed mode)\n",
      "  Starting server from /home/ubuntu/anaconda3/lib/python3.6/site-packages/h2o/backend/bin/h2o.jar\n",
      "  Ice root: /tmp/tmpz4g9f12i\n",
      "  JVM stdout: /tmp/tmpz4g9f12i/h2o_ubuntu_started_from_python.out\n",
      "  JVM stderr: /tmp/tmpz4g9f12i/h2o_ubuntu_started_from_python.err\n",
      "  Server is running at http://127.0.0.1:54321\n",
      "Connecting to H2O server at http://127.0.0.1:54321... successful.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div style=\"overflow:auto\"><table style=\"width:50%\"><tr><td>H2O cluster uptime:</td>\n",
       "<td>01 secs</td></tr>\n",
       "<tr><td>H2O cluster timezone:</td>\n",
       "<td>Etc/UTC</td></tr>\n",
       "<tr><td>H2O data parsing timezone:</td>\n",
       "<td>UTC</td></tr>\n",
       "<tr><td>H2O cluster version:</td>\n",
       "<td>3.20.0.1</td></tr>\n",
       "<tr><td>H2O cluster version age:</td>\n",
       "<td>1 month and 23 days </td></tr>\n",
       "<tr><td>H2O cluster name:</td>\n",
       "<td>H2O_from_python_ubuntu_xk3xwz</td></tr>\n",
       "<tr><td>H2O cluster total nodes:</td>\n",
       "<td>1</td></tr>\n",
       "<tr><td>H2O cluster free memory:</td>\n",
       "<td>44.44 Gb</td></tr>\n",
       "<tr><td>H2O cluster total cores:</td>\n",
       "<td>16</td></tr>\n",
       "<tr><td>H2O cluster allowed cores:</td>\n",
       "<td>16</td></tr>\n",
       "<tr><td>H2O cluster status:</td>\n",
       "<td>accepting new members, healthy</td></tr>\n",
       "<tr><td>H2O connection url:</td>\n",
       "<td>http://127.0.0.1:54321</td></tr>\n",
       "<tr><td>H2O connection proxy:</td>\n",
       "<td>None</td></tr>\n",
       "<tr><td>H2O internal security:</td>\n",
       "<td>False</td></tr>\n",
       "<tr><td>H2O API Extensions:</td>\n",
       "<td>XGBoost, Algos, AutoML, Core V3, Core V4</td></tr>\n",
       "<tr><td>Python version:</td>\n",
       "<td>3.6.5 final</td></tr></table></div>"
      ],
      "text/plain": [
       "--------------------------  ----------------------------------------\n",
       "H2O cluster uptime:         01 secs\n",
       "H2O cluster timezone:       Etc/UTC\n",
       "H2O data parsing timezone:  UTC\n",
       "H2O cluster version:        3.20.0.1\n",
       "H2O cluster version age:    1 month and 23 days\n",
       "H2O cluster name:           H2O_from_python_ubuntu_xk3xwz\n",
       "H2O cluster total nodes:    1\n",
       "H2O cluster free memory:    44.44 Gb\n",
       "H2O cluster total cores:    16\n",
       "H2O cluster allowed cores:  16\n",
       "H2O cluster status:         accepting new members, healthy\n",
       "H2O connection url:         http://127.0.0.1:54321\n",
       "H2O connection proxy:\n",
       "H2O internal security:      False\n",
       "H2O API Extensions:         XGBoost, Algos, AutoML, Core V3, Core V4\n",
       "Python version:             3.6.5 final\n",
       "--------------------------  ----------------------------------------"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# initialize H2O cluster\n",
    "h2o.init(nthreads=-1, max_mem_size='50G')\n",
    "h2o.remove_all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parse progress: |█████████████████████████████████████████████████████████| 100%\n",
      "Parse progress: |█████████████████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# transform to H2O Frame, and make sure the target variable is categorical\n",
    "h2o_train = H2OFrame(train[features + ['target']])\n",
    "h2o_test = H2OFrame(test[features])\n",
    "\n",
    "# transform into categorical\n",
    "h2o_train['target'] = h2o_train['target'].asfactor()\n",
    "\n",
    "for name in cat_features:\n",
    "    h2o_train[name] = h2o_train[name].asfactor()\n",
    "    h2o_test[name] = h2o_test[name].asfactor()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "glm Grid Build progress: |████████████████████████████████████████████████| 100%\n",
      "Best alpha:\t [0.6]\n",
      "Best lambda:\t [1.0E-6]\n",
      "Best AUC:\t 0.7231989472910549\n"
     ]
    }
   ],
   "source": [
    "# create GLM model with 5-folder cross-validation\n",
    "glm = H2OGeneralizedLinearEstimator(family='binomial', early_stopping=True, nfolds=5, \n",
    "                                    balance_classes=True, custom_metric_func='auc',\n",
    "                                    keep_cross_validation_predictions=True, seed=42)\n",
    "\n",
    "# train logistic regression model using grid search\n",
    "hyper_parameters = {'alpha': [0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0], \n",
    "                    'lambda': [0, 1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1, 1]}\n",
    "\n",
    "# grid search\n",
    "glm_grid = H2OGridSearch(glm, hyper_parameters)\n",
    "glm_grid.train(x=features, y='target', training_frame=h2o_train)\n",
    "\n",
    "# get the grid search result, sorted by AUC decreasing\n",
    "sorted_glm_grid = glm_grid.get_grid(sort_by='auc', decreasing=True)\n",
    "\n",
    "best_alpha = sorted_glm_grid.sorted_metric_table()['alpha'][0]\n",
    "best_lambda = sorted_glm_grid.sorted_metric_table()['lambda'][0]\n",
    "best_glm_auc = sorted_glm_grid.sorted_metric_table()['auc'][0]\n",
    "\n",
    "print('Best alpha:\\t', best_alpha)\n",
    "print('Best lambda:\\t', best_lambda)\n",
    "print('Best AUC:\\t', best_glm_auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "glm Model Build progress: |███████████████████████████████████████████████| 100%\n",
      "glm prediction progress: |████████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# re-build the logistic regression model with best parameters\n",
    "logit = H2OGeneralizedLinearEstimator(family='binomial', balance_classes=True, \n",
    "                                      alpha=0.6, lambda_=1.0E-6, seed=42)\n",
    "\n",
    "logit.train(x=features, y='target', training_frame=h2o_train)\n",
    "\n",
    "# make prediction\n",
    "logit_pred = logit.predict(h2o_test).as_data_frame()['p1'].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# build the ROC curve\n",
    "logit_fpr, logit_tpr, _ = roc_curve(test['target'].values, logit_pred)\n",
    "logit_auc = np.round(auc(logit_fpr, logit_tpr), 4)\n",
    "np.save('./result/h2o_logistic_fpr_over_sampling.npy', logit_fpr)\n",
    "np.save('./result/h2o_logistic_tpr_over_sampling.npy', logit_tpr)\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(8, 6))\n",
    "ax.plot(logit_fpr, logit_tpr, label='AUC: ' + str(logit_auc))\n",
    "ax.plot(logit_fpr, logit_fpr, 'k:')\n",
    "ax.set_xlabel('False Positive Rate', fontsize=12)\n",
    "ax.set_ylabel('True Positive Rate', fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Random Forest\n",
    "Since it's too slow to run this module, the number of trees `ntrees` are only set to be $100$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "drf Grid Build progress: |████████████████████████████████████████████████| 100%\n",
      "Best max_depth:\t\t 15\n",
      "Best mtries:\t\t 20\n",
      "Best sample_rate:\t 0.632\n",
      "Best AUC:\t\t 0.722763705793492\n"
     ]
    }
   ],
   "source": [
    "# random forest grid search\n",
    "hyper_parameters = {'max_depth': [15, 20, 25, 30, 35], \n",
    "                    'mtries': [-1, 8, 15, 20, 25], \n",
    "                    'sample_rate': [0.5, 0.632, 0.75]}\n",
    "\n",
    "# build random forest model\n",
    "rf_model = H2ORandomForestEstimator(balance_classes=True, ntrees=100, nfolds=5, \n",
    "                                    stopping_rounds=5, stopping_metric='auc', \n",
    "                                    keep_cross_validation_predictions=True)\n",
    "\n",
    "# define search criteria\n",
    "search_criteria = {'strategy': 'RandomDiscrete', 'max_models': 10, 'seed': 42, \n",
    "                   'stopping_metric': 'auc', 'stopping_rounds': 5}\n",
    "\n",
    "# grid search\n",
    "rf_grid = H2OGridSearch(rf_model, hyper_parameters, search_criteria=search_criteria)\n",
    "\n",
    "rf_grid.train(x=features, y='target', training_frame=h2o_train)\n",
    "\n",
    "# get the grid search result, sorted by AUC decreasing\n",
    "sorted_rf_grid = rf_grid.get_grid(sort_by='auc', decreasing=True)\n",
    "\n",
    "best_max_depth = sorted_rf_grid.sorted_metric_table()['max_depth'][0]\n",
    "best_mtries = sorted_rf_grid.sorted_metric_table()['mtries'][0]\n",
    "best_sample_rate = sorted_rf_grid.sorted_metric_table()['sample_rate'][0]\n",
    "best_rf_auc = sorted_rf_grid.sorted_metric_table()['auc'][0]\n",
    "\n",
    "print('Best max_depth:\\t\\t', best_max_depth)\n",
    "print('Best mtries:\\t\\t', best_mtries)\n",
    "print('Best sample_rate:\\t', best_sample_rate)\n",
    "print('Best AUC:\\t\\t', best_rf_auc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "drf Model Build progress: |███████████████████████████████████████████████| 100%\n",
      "drf prediction progress: |████████████████████████████████████████████████| 100%\n"
     ]
    }
   ],
   "source": [
    "# build random forest model with best parameters\n",
    "h2o_rf = H2ORandomForestEstimator(balance_classes=True, max_depth=15, mtries=20, \n",
    "                                  sample_rate=0.632, ntrees=1000, seed=42)\n",
    "\n",
    "h2o_rf.train(x=features, y='target', training_frame=h2o_train)\n",
    "\n",
    "# make prediction\n",
    "h2o_rf_pred = h2o_rf.predict(h2o_test).as_data_frame()['p1'].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# get feature importance details\n",
    "h2o_rf_importance = h2o_rf.varimp(use_pandas=True)\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.barplot(x='scaled_importance', y='variable', data=h2o_rf_importance[:15], ax=ax)\n",
    "ax.set_title('LightGBM Random Forest Feature Importance', fontsize=16)\n",
    "ax.set_xlabel('Relative Importance', fontsize=12)\n",
    "ax.set_ylabel('')\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# build the ROC curve\n",
    "h2o_rf_fpr, h2o_rf_tpr, _ = roc_curve(test['target'].values, h2o_rf_pred)\n",
    "h2o_rf_auc = np.round(auc(h2o_rf_fpr, h2o_rf_tpr), 4)\n",
    "np.save('./result/h2o_rf_fpr_over_sampling.npy', h2o_rf_fpr)\n",
    "np.save('./result/h2o_rf_tpr_over_sampling.npy', h2o_rf_tpr)\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(8, 6))\n",
    "ax.plot(h2o_rf_fpr, h2o_rf_tpr, label='AUC: ' + str(h2o_rf_auc))\n",
    "ax.plot(h2o_rf_fpr, h2o_rf_fpr, 'k:')\n",
    "ax.set_xlabel('False Positive Rate', fontsize=12)\n",
    "ax.set_ylabel('True Positive Rate', fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "H2O session _sid_9a45 closed.\n"
     ]
    }
   ],
   "source": [
    "# Shutdown h2o instance\n",
    "h2o.cluster().shutdown()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# II. LightGBM Data Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# encoding categorical data into numerical format\n",
    "label_encoders = []\n",
    "for name in cat_features:\n",
    "    encoder = LabelEncoder()\n",
    "    train[name] = encoder.fit_transform(train[name])\n",
    "    test[name] = encoder.transform(test[name])\n",
    "    label_encoders.append(encoder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create LightGBM dataset\n",
    "train_x = train[features]\n",
    "train_y = train['target'].values\n",
    "\n",
    "gbm_train = lgb.Dataset(data=train_x, label=train_y, feature_name=features, \n",
    "                        categorical_feature=cat_features, free_raw_data=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define parameter space to explore\n",
    "rf_num_leaves_list = [20, 30, 40, 50, 60]\n",
    "rf_max_depth_list = [-1, 20, 30, 40, 50]\n",
    "rf_min_data_in_leaf_list = [20, 30, 40, 50]\n",
    "rf_bagging_frac_list = [0.5, 0.632, 0.7, 0.8]\n",
    "rf_feature_frac_list = [0.4, 0.5, 0.6, 0.7, 0.8]\n",
    "\n",
    "rf_num_leaves_vals = []\n",
    "rf_max_depth_vals = []\n",
    "rf_min_data_vals = []\n",
    "rf_bagging_frac_vals = []\n",
    "rf_feature_frac_vals = []\n",
    "\n",
    "rf_mean_auc = []\n",
    "rf_std_auc = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|num_leaves |max_depth |min_data_in_leaf |bagging_fraction |feature_fraction | AUC   | std   |\n",
      "----------------------------------------------------------------------------------------------\n",
      "|        50 |       50 |              40 |          0.7000 |          0.8000 |0.7126 |0.0013 |\n",
      "|        60 |       20 |              40 |          0.7000 |          0.6000 |0.7148 |0.0012 |\n",
      "|        40 |       50 |              50 |          0.8000 |          0.6000 |0.7125 |0.0011 |\n",
      "|        60 |       20 |              50 |          0.8000 |          0.5000 |0.7155 |0.0011 |\n",
      "|        50 |       50 |              20 |          0.8000 |          0.5000 |0.7146 |0.0011 |\n",
      "|        60 |       40 |              20 |          0.5000 |          0.6000 |0.7151 |0.0011 |\n",
      "|        40 |       20 |              50 |          0.8000 |          0.7000 |0.7118 |0.0011 |\n",
      "|        40 |       40 |              40 |          0.8000 |          0.4000 |0.7148 |0.0011 |\n",
      "|        40 |       50 |              40 |          0.7000 |          0.8000 |0.7114 |0.0012 |\n",
      "|        20 |       20 |              50 |          0.5000 |          0.7000 |0.7081 |0.0010 |\n",
      "|        30 |       20 |              20 |          0.6320 |          0.8000 |0.7097 |0.0011 |\n",
      "|        30 |       40 |              50 |          0.7000 |          0.7000 |0.7102 |0.0011 |\n",
      "|        50 |       50 |              50 |          0.7000 |          0.6000 |0.7139 |0.0011 |\n",
      "|        20 |       40 |              30 |          0.8000 |          0.7000 |0.7075 |0.0011 |\n",
      "|        30 |       20 |              50 |          0.6320 |          0.8000 |0.7097 |0.0011 |\n",
      "|        30 |       20 |              50 |          0.6320 |          0.5000 |0.7121 |0.0010 |\n",
      "|        50 |       40 |              20 |          0.6320 |          0.8000 |0.7128 |0.0012 |\n",
      "|        60 |       20 |              40 |          0.5000 |          0.5000 |0.7159 |0.0011 |\n",
      "|        20 |       40 |              50 |          0.8000 |          0.8000 |0.7068 |0.0012 |\n",
      "|        20 |       50 |              40 |          0.5000 |          0.4000 |0.7112 |0.0009 |\n",
      "|        20 |       -1 |              50 |          0.5000 |          0.7000 |0.7081 |0.0010 |\n",
      "|        40 |       30 |              20 |          0.8000 |          0.6000 |0.7125 |0.0011 |\n",
      "|        40 |       -1 |              40 |          0.5000 |          0.5000 |0.7139 |0.0010 |\n",
      "|        30 |       -1 |              50 |          0.7000 |          0.4000 |0.7133 |0.0010 |\n",
      "|        50 |       20 |              20 |          0.8000 |          0.8000 |0.7122 |0.0013 |\n",
      "|        40 |       40 |              50 |          0.6320 |          0.6000 |0.7128 |0.0011 |\n",
      "|        40 |       -1 |              40 |          0.5000 |          0.6000 |0.7131 |0.0011 |\n",
      "|        20 |       50 |              30 |          0.7000 |          0.6000 |0.7084 |0.0010 |\n",
      "|        20 |       20 |              30 |          0.8000 |          0.8000 |0.7068 |0.0012 |\n",
      "|        40 |       -1 |              50 |          0.5000 |          0.7000 |0.7125 |0.0011 |\n",
      "----------------------------------------------------------------------------------------------\n",
      "|        60 |       20 |              40 |          0.5000 |          0.5000 |0.7159 |0.0011 |\n"
     ]
    }
   ],
   "source": [
    "# Random search with Cross validation\n",
    "s = '|{0:^10s} |{1:^9s} |{2:^16s} |{3:^16s} |{4:^16s} |{5:^6s} |{6:^6s} |'\n",
    "print(s.format('num_leaves', 'max_depth', 'min_data_in_leaf', 'bagging_fraction', \n",
    "               'feature_fraction', 'AUC', 'std'))\n",
    "print('-' * 94)\n",
    "\n",
    "# perform random search for given number n\n",
    "n = 30\n",
    "np.random.seed(42)\n",
    "visited = set()\n",
    "for i in range(n):\n",
    "    while True:\n",
    "        num_leaves = np.random.choice(rf_num_leaves_list)\n",
    "        max_depth = np.random.choice(rf_max_depth_list)\n",
    "        min_data_in_leaf = np.random.choice(rf_min_data_in_leaf_list)\n",
    "        bagging_fraction = np.random.choice(rf_bagging_frac_list)\n",
    "        feature_fraction = np.random.choice(rf_feature_frac_list)\n",
    "        \n",
    "        tuples = (num_leaves, max_depth, min_data_in_leaf, bagging_fraction, feature_fraction)\n",
    "        if tuples not in visited:\n",
    "            visited.add(tuples)\n",
    "            break\n",
    "            \n",
    "    params = {'objective': 'binary', \n",
    "              'boosting': 'rf', \n",
    "              'num_threads': 16, \n",
    "              'is_unbalance': True, \n",
    "              'metric': ['auc'],\n",
    "              'learning_rate': 0.1, \n",
    "              'max_bin': 255, \n",
    "              'num_leaves': num_leaves, \n",
    "              'max_depth': max_depth, \n",
    "              'min_data_in_leaf': min_data_in_leaf, \n",
    "              'bagging_fraction': bagging_fraction, \n",
    "              'feature_fraction': feature_fraction, \n",
    "              'bagging_freq': 1, \n",
    "              'lambda_l1': 0.0,\n",
    "              'lambda_l2': 0.0,\n",
    "              'drop_rate': 0.1, \n",
    "              'seed': 42}\n",
    "\n",
    "    # 5-folder cross validation (no early stopping)\n",
    "    history = lgb.cv(params, train_set=gbm_train, nfold=5, num_boost_round=1000, \n",
    "                     stratified=True, early_stopping_rounds=None, verbose_eval=False, \n",
    "                     seed=42, feature_name=features, categorical_feature=cat_features)\n",
    "\n",
    "    # get result\n",
    "    rf_num_leaves_vals.append(num_leaves)\n",
    "    rf_max_depth_vals.append(max_depth)\n",
    "    rf_min_data_vals.append(min_data_in_leaf)\n",
    "    rf_bagging_frac_vals.append(bagging_fraction)\n",
    "    rf_feature_frac_vals.append(feature_fraction)\n",
    "    rf_mean_auc.append(history['auc-mean'][-1])\n",
    "    rf_std_auc.append(history['auc-stdv'][-1])\n",
    "        \n",
    "    # output the resuts\n",
    "    ss = '|{0:10d} |{1:9d} |{2:16d} |{3:16.4f} |{4:16.4f} |{5:6.4f} |{6:6.4f} |'\n",
    "    print(ss.format(num_leaves, max_depth, min_data_in_leaf, bagging_fraction, \n",
    "                    feature_fraction, history['auc-mean'][-1], history['auc-stdv'][-1]))\n",
    "    \n",
    "# get the best parameters\n",
    "idx = np.argmax(rf_mean_auc)\n",
    "print('-' * 94)\n",
    "print(ss.format(rf_num_leaves_vals[idx], rf_max_depth_vals[idx], rf_min_data_vals[idx], \n",
    "                rf_bagging_frac_vals[idx], rf_feature_frac_vals[idx], rf_mean_auc[idx], \n",
    "                rf_std_auc[idx]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define best parameters\n",
    "params = {'objective': 'binary', \n",
    "          'boosting': 'rf', \n",
    "          'num_threads': 16, \n",
    "          'is_unbalance': True, \n",
    "          'metric': ['auc'],\n",
    "          'learning_rate': 0.1, \n",
    "          'max_bin': 255, \n",
    "          'num_leaves': 60, \n",
    "          'max_depth': 20, \n",
    "          'min_data_in_leaf': 40, \n",
    "          'bagging_fraction': 0.5, \n",
    "          'feature_fraction': 0.5, \n",
    "          'bagging_freq': 1, \n",
    "          'lambda_l1': 0.0,\n",
    "          'lambda_l2': 0.0,\n",
    "          'drop_rate': 0.1, \n",
    "          'seed': 42}\n",
    "\n",
    "# re-train the model and make predictions\n",
    "lgb_rf = lgb.train(params, train_set=gbm_train, num_boost_round=1000, \n",
    "                   feature_name=features, categorical_feature=cat_features)\n",
    "\n",
    "lgb_rf_pred = lgb_rf.predict(test[features])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# get feature importance details\n",
    "importance = lgb_rf.feature_importance()\n",
    "lgb_rf_importance = pd.DataFrame({'feature': features, 'importance': importance}, \n",
    "                                 columns=['feature', 'importance'])\n",
    "lgb_rf_importance = lgb_rf_importance.sort_values(by='importance', ascending=False)\n",
    "lgb_rf_importance['importance'] = lgb_rf_importance['importance'] / lgb_rf_importance['importance'].max()\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.barplot(x='importance', y='feature', data=lgb_rf_importance[:15], ax=ax)\n",
    "ax.set_title('LightGBM Random Forest Feature Importance', fontsize=16)\n",
    "ax.set_xlabel('Relative Importance', fontsize=12)\n",
    "ax.set_ylabel('')\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# build the ROC curve\n",
    "lgb_rf_fpr, lgb_rf_tpr, _ = roc_curve(test['target'].values, lgb_rf_pred)\n",
    "lgb_rf_auc = np.round(auc(lgb_rf_fpr, lgb_rf_tpr), 4)\n",
    "np.save('./result/lgb_rf_fpr_over_sampling.npy', lgb_rf_fpr)\n",
    "np.save('./result/lgb_rf_tpr_over_sampling.npy', lgb_rf_tpr)\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(8, 6))\n",
    "ax.plot(lgb_rf_fpr, lgb_rf_tpr, label='AUC: ' + str(lgb_rf_auc))\n",
    "ax.plot(lgb_rf_fpr, lgb_rf_fpr, 'k:')\n",
    "ax.set_xlabel('False Positive Rate', fontsize=12)\n",
    "ax.set_ylabel('True Positive Rate', fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Boosting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define parameter space to explore\n",
    "gbm_learning_rate_list = [0.03, 0.05, 0.1]\n",
    "gbm_num_leaves_list = [20, 30, 40, 50, 60]\n",
    "gbm_max_depth_list = [-1, 10, 20, 30, 40]\n",
    "gbm_min_data_in_leaf_list = [20, 30, 40, 50]\n",
    "\n",
    "gbm_learning_rate_vals = []\n",
    "gbm_num_leaves_vals = []\n",
    "gbm_max_depth_vals = []\n",
    "gbm_min_data_vals = []\n",
    "\n",
    "gbm_best_rounds = []\n",
    "gbm_mean_auc = []\n",
    "gbm_std_auc = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "| learning_rate | num_leaves | max_depth  | min_data_in_leaf | best_rounds |  AUC   |  std   |\n",
      "----------------------------------------------------------------------------------------------\n",
      "|       0.10000 |         50 |         40 |               40 |          219| 0.7371 | 0.0016 |\n",
      "|       0.10000 |         60 |         40 |               40 |          180| 0.7368 | 0.0016 |\n",
      "|       0.05000 |         40 |         20 |               40 |          668| 0.7383 | 0.0015 |\n",
      "|       0.03000 |         50 |         20 |               30 |          848| 0.7382 | 0.0014 |\n",
      "|       0.03000 |         30 |         30 |               30 |         1000| 0.7382 | 0.0014 |\n",
      "|       0.05000 |         30 |         30 |               20 |          681| 0.7381 | 0.0014 |\n",
      "|       0.03000 |         50 |         10 |               30 |          799| 0.7382 | 0.0014 |\n",
      "|       0.03000 |         50 |         -1 |               20 |          972| 0.7383 | 0.0014 |\n",
      "|       0.10000 |         40 |         10 |               50 |          305| 0.7376 | 0.0014 |\n",
      "|       0.10000 |         40 |         30 |               40 |          240| 0.7372 | 0.0016 |\n",
      "|       0.03000 |         40 |         40 |               40 |          999| 0.7384 | 0.0015 |\n",
      "|       0.10000 |         60 |         -1 |               40 |          180| 0.7368 | 0.0016 |\n",
      "|       0.05000 |         50 |         -1 |               50 |          570| 0.7383 | 0.0016 |\n",
      "|       0.05000 |         30 |         10 |               20 |          598| 0.7380 | 0.0015 |\n",
      "|       0.05000 |         60 |         10 |               50 |          556| 0.7381 | 0.0015 |\n",
      "|       0.10000 |         50 |         30 |               20 |          239| 0.7370 | 0.0016 |\n",
      "|       0.10000 |         40 |         -1 |               50 |          300| 0.7373 | 0.0015 |\n",
      "|       0.05000 |         50 |         10 |               30 |          540| 0.7381 | 0.0015 |\n",
      "|       0.05000 |         30 |         30 |               30 |          735| 0.7382 | 0.0015 |\n",
      "|       0.03000 |         30 |         10 |               50 |         1000| 0.7383 | 0.0014 |\n",
      "|       0.10000 |         50 |         -1 |               30 |          191| 0.7369 | 0.0014 |\n",
      "|       0.03000 |         60 |         10 |               40 |          628| 0.7381 | 0.0015 |\n",
      "|       0.03000 |         30 |         -1 |               50 |          999| 0.7384 | 0.0015 |\n",
      "|       0.03000 |         20 |         40 |               40 |          999| 0.7376 | 0.0014 |\n",
      "|       0.03000 |         20 |         -1 |               40 |          999| 0.7376 | 0.0014 |\n",
      "|       0.03000 |         20 |         30 |               50 |         1000| 0.7376 | 0.0013 |\n",
      "|       0.10000 |         40 |         20 |               20 |          249| 0.7373 | 0.0015 |\n",
      "|       0.10000 |         40 |         -1 |               40 |          240| 0.7372 | 0.0016 |\n",
      "|       0.03000 |         30 |         10 |               20 |         1000| 0.7382 | 0.0014 |\n",
      "|       0.10000 |         20 |         30 |               30 |          434| 0.7376 | 0.0015 |\n",
      "----------------------------------------------------------------------------------------------\n",
      "|       0.03000 |         40 |         40 |               40 |          999| 0.7384 | 0.0015 |\n"
     ]
    }
   ],
   "source": [
    "# Random search with Cross validation\n",
    "s = '| {0:^3s} | {1:^10s} | {2:^10s} | {3:^16s} | {4:^12s}| {5:^6s} | {6:^6s} |'\n",
    "print(s.format('learning_rate', 'num_leaves', 'max_depth', 'min_data_in_leaf', \n",
    "               'best_rounds', 'AUC', 'std'))\n",
    "print('-' * 94)\n",
    "\n",
    "# perform random search for given number n\n",
    "n = 30\n",
    "np.random.seed(42)\n",
    "visited = set()\n",
    "for i in range(n):\n",
    "    while True:\n",
    "        learning_rate = np.random.choice(gbm_learning_rate_list)\n",
    "        num_leaves = np.random.choice(gbm_num_leaves_list)\n",
    "        max_depth = np.random.choice(gbm_max_depth_list)\n",
    "        min_data_in_leaf = np.random.choice(gbm_min_data_in_leaf_list)\n",
    "        \n",
    "        tuples = (learning_rate, num_leaves, max_depth, min_data_in_leaf)\n",
    "        if tuples not in visited:\n",
    "            visited.add(tuples)\n",
    "            break\n",
    "            \n",
    "    params = {'objective': 'binary', \n",
    "              'boosting': 'gbdt', \n",
    "              'num_threads': 16, \n",
    "              'is_unbalance': True, \n",
    "              'metric': ['auc'],\n",
    "              'max_bin': 255, \n",
    "              'learning_rate': learning_rate, \n",
    "              'num_leaves': num_leaves, \n",
    "              'max_depth': max_depth, \n",
    "              'min_data_in_leaf': min_data_in_leaf, \n",
    "              'bagging_fraction': 1.0, \n",
    "              'feature_fraction': 1.0, \n",
    "              'bagging_freq': 0, \n",
    "              'lambda_l1': 0.0,\n",
    "              'lambda_l2': 0.0,\n",
    "              'drop_rate': 0.1, \n",
    "              'seed': 42}\n",
    "\n",
    "    # 5-folder cross validation (no early stopping)\n",
    "    history = lgb.cv(params, train_set=gbm_train, nfold=5, num_boost_round=1000, \n",
    "                     stratified=True, early_stopping_rounds=10, verbose_eval=False, \n",
    "                     seed=42, feature_name=features, categorical_feature=cat_features)\n",
    "\n",
    "    # get result\n",
    "    gbm_learning_rate_vals.append(learning_rate)\n",
    "    gbm_num_leaves_vals.append(num_leaves)\n",
    "    gbm_max_depth_vals.append(max_depth)\n",
    "    gbm_min_data_vals.append(min_data_in_leaf)\n",
    "    gbm_best_rounds.append(len(history['auc-mean']))\n",
    "    gbm_mean_auc.append(history['auc-mean'][-1])\n",
    "    gbm_std_auc.append(history['auc-stdv'][-1])\n",
    "        \n",
    "    # output the resuts\n",
    "    ss = '| {0:>13.5f} | {1:>10d} | {2:>10d} | {3:>16d} | {4:>12d}| {5:>6.4f} | {6:>6.4f} |'\n",
    "    print(ss.format(learning_rate, num_leaves, max_depth, min_data_in_leaf, \n",
    "                    len(history['auc-mean']), history['auc-mean'][-1], \n",
    "                    history['auc-stdv'][-1]))\n",
    "    \n",
    "# get the best parameters\n",
    "idx = np.argmax(gbm_mean_auc)\n",
    "print('-' * 94)\n",
    "print(ss.format(gbm_learning_rate_vals[idx], gbm_num_leaves_vals[idx], gbm_max_depth_vals[idx], \n",
    "                gbm_min_data_vals[idx], gbm_best_rounds[idx], gbm_mean_auc[idx], gbm_std_auc[idx]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define best parameters\n",
    "params = {'objective': 'binary', \n",
    "          'boosting': 'gbdt', \n",
    "          'num_threads': 4, \n",
    "          'is_unbalance': True, \n",
    "          'metric': ['auc'],\n",
    "          'max_bin': 255, \n",
    "          'learning_rate': 0.03, \n",
    "          'num_leaves': 40, \n",
    "          'max_depth': 40, \n",
    "          'min_data_in_leaf': 40, \n",
    "          'bagging_fraction': 1.0, \n",
    "          'feature_fraction': 1.0,\n",
    "          'bagging_freq': 0,\n",
    "          'lambda_l1': 0.0,\n",
    "          'lambda_l2': 0.0,\n",
    "          'drop_rate': 0.1, \n",
    "          'seed': 42}\n",
    "\n",
    "# re-train the model and make predictions\n",
    "lgb_gbm = lgb.train(params, train_set=gbm_train, num_boost_round=999, \n",
    "                feature_name=features, categorical_feature=cat_features)\n",
    "lgb_gbm_pred = lgb_gbm.predict(test[features])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# get feature importance details\n",
    "importance = lgb_gbm.feature_importance()\n",
    "lgb_gbm_importance = pd.DataFrame({'feature': features, 'importance': importance}, \n",
    "                                  columns=['feature', 'importance'])\n",
    "lgb_gbm_importance = lgb_gbm_importance.sort_values(by='importance', ascending=False)\n",
    "lgb_gbm_importance['importance'] = lgb_gbm_importance['importance'] / lgb_gbm_importance['importance'].max()\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.barplot(x='importance', y='feature', data=lgb_gbm_importance[:15], ax=ax)\n",
    "ax.set_title('LightGBM GBM Feature Importance', fontsize=16)\n",
    "ax.set_xlabel('Relative Importance', fontsize=12)\n",
    "ax.set_ylabel('')\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# build the ROC curve\n",
    "lgb_gbm_fpr, lgb_gbm_tpr, _ = roc_curve(test['target'].values, lgb_gbm_pred)\n",
    "lgb_gbm_auc = np.round(auc(lgb_gbm_fpr, lgb_gbm_tpr), 4)\n",
    "np.save('./result/lgb_gbm_fpr_over_sampling.npy', lgb_gbm_fpr)\n",
    "np.save('./result/lgb_gbm_tpr_over_sampling.npy', lgb_gbm_tpr)\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(8, 6))\n",
    "ax.plot(lgb_gbm_fpr, lgb_gbm_tpr, label='AUC: ' + str(lgb_gbm_auc))\n",
    "ax.plot(lgb_gbm_fpr, lgb_gbm_fpr, 'k:')\n",
    "ax.set_xlabel('False Positive Rate', fontsize=12)\n",
    "ax.set_ylabel('True Positive Rate', fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# III. CatBoost Data Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create Pool object\n",
    "train_pool = Pool(data=train[features], label=train['target'].values, feature_names=features, \n",
    "                  cat_features=np.array(range(len(cat_features))))\n",
    "\n",
    "test_pool = Pool(data=test[features], feature_names=features, \n",
    "                 cat_features=np.array(range(len(cat_features))))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Boosting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define parameter space to explore\n",
    "learning_rate_list = [0.03, 0.05, 0.08, 0.1]\n",
    "depth_list = [4, 5, 6, 7, 8, 9, 10]\n",
    "l2_leaf_reg_list = [1, 3, 5, 7, 9]\n",
    "random_strength_list = [0.1, 0.5, 1, 2]\n",
    "bagging_temperature_list = [0, 0.2, 0.4, 0.6, 0.8, 1.0]\n",
    "\n",
    "learning_rate_values = []\n",
    "depth_values = []\n",
    "l2_leaf_reg_values = []\n",
    "random_strength_values = []\n",
    "bagging_temperature_values = []\n",
    "\n",
    "best_iterations_values = []\n",
    "train_mean_auc_values = []\n",
    "test_mean_auc_values = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "|learning_rate |depth |l2_leaf_reg |strength |temperature |iterations |train_AUC | test_AUC |\n",
      "---------------------------------------------------------------------------------------------\n",
      "|       0.0800 |    7 |          9 |  1.0000 |     0.4000 |       797 |   0.7673 |   0.7400 |\n",
      "|       0.1000 |    8 |          9 |  1.0000 |     0.2000 |       442 |   0.7709 |   0.7390 |\n",
      "|       0.0800 |   10 |          5 |  1.0000 |     0.8000 |       372 |   0.7962 |   0.7368 |\n",
      "|       0.1000 |    6 |          9 |  0.5000 |     0.6000 |       863 |   0.7611 |   0.7398 |\n",
      "|       0.0500 |    9 |          3 |  2.0000 |     0.6000 |       787 |   0.7881 |   0.7386 |\n",
      "|       0.0300 |    4 |          7 |  0.5000 |     1.0000 |      1000 |   0.7358 |   0.7331 |\n",
      "|       0.0300 |    7 |          1 |  0.1000 |     0.4000 |      1000 |   0.7529 |   0.7378 |\n",
      "|       0.0800 |   10 |          3 |  2.0000 |     0.6000 |       301 |   0.7891 |   0.7363 |\n",
      "|       0.1000 |   10 |          5 |  2.0000 |     0.6000 |       279 |   0.7931 |   0.7363 |\n",
      "|       0.1000 |    4 |          5 |  0.1000 |     0.4000 |      1000 |   0.7486 |   0.7390 |\n",
      "|       0.0800 |    8 |          1 |  1.0000 |     0.2000 |       544 |   0.7752 |   0.7387 |\n",
      "|       0.1000 |    4 |          7 |  0.5000 |     0.2000 |      1000 |   0.7491 |   0.7390 |\n",
      "|       0.0500 |    4 |          3 |  0.1000 |     0.2000 |      1000 |   0.7418 |   0.7365 |\n",
      "|       0.1000 |    7 |          7 |  1.0000 |     0.6000 |       696 |   0.7684 |   0.7397 |\n",
      "|       0.0300 |   10 |          5 |  0.5000 |     0.0000 |      1000 |   0.7984 |   0.7387 |\n",
      "|       0.1000 |    5 |          7 |  0.5000 |     1.0000 |       911 |   0.7510 |   0.7389 |\n",
      "|       0.0500 |    9 |          3 |  2.0000 |     1.0000 |       725 |   0.7824 |   0.7381 |\n",
      "|       0.0300 |   10 |          3 |  0.5000 |     0.6000 |       948 |   0.7991 |   0.7381 |\n",
      "|       0.0500 |    5 |          7 |  0.5000 |     1.0000 |      1000 |   0.7442 |   0.7372 |\n",
      "|       0.0800 |    7 |          1 |  0.5000 |     0.8000 |       757 |   0.7659 |   0.7392 |\n",
      "|       0.1000 |    8 |          3 |  1.0000 |     0.8000 |       491 |   0.7737 |   0.7384 |\n",
      "|       0.1000 |    5 |          1 |  2.0000 |     0.6000 |       864 |   0.7531 |   0.7394 |\n",
      "|       0.1000 |    8 |          1 |  0.1000 |     0.8000 |       417 |   0.7703 |   0.7376 |\n",
      "|       0.0300 |    4 |          1 |  2.0000 |     0.0000 |      1000 |   0.7371 |   0.7338 |\n",
      "|       0.1000 |   10 |          5 |  1.0000 |     0.0000 |       257 |   0.7903 |   0.7368 |\n",
      "|       0.1000 |    6 |          5 |  0.1000 |     0.4000 |       722 |   0.7593 |   0.7396 |\n",
      "|       0.0300 |    5 |          3 |  0.1000 |     0.6000 |      1000 |   0.7403 |   0.7354 |\n",
      "|       0.0800 |    4 |          7 |  0.5000 |     0.0000 |      1000 |   0.7468 |   0.7384 |\n",
      "|       0.1000 |   10 |          9 |  1.0000 |     0.6000 |       333 |   0.7984 |   0.7370 |\n",
      "|       0.1000 |    9 |          5 |  1.0000 |     0.0000 |       329 |   0.7792 |   0.7380 |\n",
      "---------------------------------------------------------------------------------------------\n",
      "|       0.0800 |    7 |          9 |  1.0000 |     0.4000 |       797 |   0.7673 |   0.7400 |\n"
     ]
    }
   ],
   "source": [
    "# Random search with Cross validation\n",
    "s = '|{0:>13s} |{1:>5s} |{2:>11s} |{3:>8s} |{4:>11s} |{5:>10s} |{6:>9s} |{7:>9s} |'\n",
    "print(s.format('learning_rate', 'depth', 'l2_leaf_reg', 'strength', 'temperature', \n",
    "               'iterations', 'train_AUC', 'test_AUC'))\n",
    "print('-' * 93)\n",
    "\n",
    "# perform random search for given number n\n",
    "n = 30\n",
    "np.random.seed(42)\n",
    "visited = set()\n",
    "for i in range(n):\n",
    "    while True:\n",
    "        learning_rate = np.random.choice(learning_rate_list)\n",
    "        depth = np.random.choice(depth_list)\n",
    "        l2_leaf_reg = np.random.choice(l2_leaf_reg_list)\n",
    "        random_strength = np.random.choice(random_strength_list)\n",
    "        bagging_temperature = np.random.choice(bagging_temperature_list)\n",
    "        \n",
    "        tuples = (learning_rate, depth, l2_leaf_reg, random_strength, bagging_temperature)\n",
    "        if tuples not in visited:\n",
    "            visited.add(tuples)\n",
    "            break\n",
    "            \n",
    "    # define parameters\n",
    "    params = {'loss_function': 'Logloss', \n",
    "              'custom_metric': 'AUC', \n",
    "              'eval_metric': 'AUC',\n",
    "              'learning_rate': learning_rate, \n",
    "              'depth': depth, \n",
    "              'l2_leaf_reg': l2_leaf_reg,\n",
    "              'random_strength': random_strength, \n",
    "              'bagging_temperature': bagging_temperature, \n",
    "              'random_seed': 42, \n",
    "              'bootstrap_type': 'Bayesian', \n",
    "              'has_time': False, \n",
    "              'class_weights': [1, 3.2]}\n",
    "\n",
    "    scores = cv(pool=train_pool, params=params, iterations=1000, fold_count=5,\n",
    "                seed=42, shuffle=True, logging_level='Silent', stratified=True,\n",
    "                as_pandas=False, metric_period=1, early_stopping_rounds=5)\n",
    "    \n",
    "    # get result\n",
    "    learning_rate_values.append(learning_rate)\n",
    "    depth_values.append(depth)\n",
    "    l2_leaf_reg_values.append(l2_leaf_reg)\n",
    "    random_strength_values.append(random_strength)\n",
    "    bagging_temperature_values.append(bagging_temperature)\n",
    "    \n",
    "    best_idx = np.argmax(scores['test-AUC-mean'])\n",
    "    best_iterations = best_idx + 1\n",
    "    train_mean_auc = scores['train-AUC-mean'][best_idx]\n",
    "    test_mean_auc = scores['test-AUC-mean'][best_idx]\n",
    "    \n",
    "    best_iterations_values.append(best_iterations)\n",
    "    train_mean_auc_values.append(train_mean_auc)\n",
    "    test_mean_auc_values.append(test_mean_auc)\n",
    "    \n",
    "    # output the resuts\n",
    "    ss = '|{0:>13.4f} |{1:>5d} |{2:>11d} |{3:>8.4f} |{4:>11.4f} |{5:>10d} |{6:>9.4f} |{7:>9.4f} |'\n",
    "    print(ss.format(learning_rate, depth, l2_leaf_reg, random_strength, bagging_temperature, \n",
    "                    best_iterations, train_mean_auc, test_mean_auc))\n",
    "    \n",
    "# get the best parameters\n",
    "idx = np.argmax(test_mean_auc_values)\n",
    "print('-' * 93)\n",
    "print(ss.format(learning_rate_values[idx], depth_values[idx], l2_leaf_reg_values[idx], \n",
    "                random_strength_values[idx], bagging_temperature_values[idx], \n",
    "                best_iterations_values[idx], train_mean_auc_values[idx], \n",
    "                test_mean_auc_values[idx]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0:\ttotal: 408ms\tremaining: 5m 24s\n",
      "100:\ttotal: 38.3s\tremaining: 4m 24s\n",
      "200:\ttotal: 1m 15s\tremaining: 3m 45s\n",
      "300:\ttotal: 1m 53s\tremaining: 3m 7s\n",
      "400:\ttotal: 2m 31s\tremaining: 2m 29s\n",
      "500:\ttotal: 3m 9s\tremaining: 1m 52s\n",
      "600:\ttotal: 3m 48s\tremaining: 1m 14s\n",
      "700:\ttotal: 4m 26s\tremaining: 36.4s\n",
      "796:\ttotal: 5m 2s\tremaining: 0us\n"
     ]
    }
   ],
   "source": [
    "# build CatBoost classifier\n",
    "cat_gbm = CatBoostClassifier(loss_function='Logloss', custom_metric='AUC', eval_metric='AUC',\n",
    "                             learning_rate=0.8, depth=7, l2_leaf_reg=9, random_strength=1, \n",
    "                             bagging_temperature=0.4, iterations=797, random_seed=42, \n",
    "                             class_weights=[1, 3.2], bootstrap_type='Bayesian')\n",
    "\n",
    "cat_gbm.fit(X=train_pool, eval_set=None, logging_level='Verbose', plot=False, \n",
    "            column_description=None, metric_period=100, early_stopping_rounds=None)\n",
    "\n",
    "cat_gbm_pred = cat_gbm.predict_proba(data=test_pool)[:, 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# get feature importance details\n",
    "importance = cat_gbm.get_feature_importance(data=None, prettified=True)\n",
    "\n",
    "names = []\n",
    "vals = []\n",
    "for name, val in importance:\n",
    "    names.append(str(name.decode('ASCII')))\n",
    "    vals.append(val)\n",
    "\n",
    "cat_gbm_importance = pd.DataFrame({'feature': names, 'importance': vals}, \n",
    "                                  columns=['feature', 'importance'])\n",
    "\n",
    "cat_gbm_importance['importance'] = cat_gbm_importance['importance'] / cat_gbm_importance['importance'].max()\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(10, 8))\n",
    "sns.barplot(x='importance', y='feature', data=cat_gbm_importance[:15], ax=ax)\n",
    "ax.set_title('CatBoost GBM Feature Importance', fontsize=16)\n",
    "ax.set_xlabel('Relative Importance', fontsize=12)\n",
    "ax.set_ylabel('')\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# build the ROC curve\n",
    "cat_gbm_fpr, cat_gbm_tpr, _ = roc_curve(test['target'].values, cat_gbm_pred)\n",
    "cat_gbm_auc = np.round(auc(cat_gbm_fpr, cat_gbm_tpr), 4)\n",
    "np.save('./result/cat_gbm_fpr_over_sampling.npy', cat_gbm_fpr)\n",
    "np.save('./result/cat_gbm_tpr_over_sampling.npy', cat_gbm_tpr)\n",
    "\n",
    "# visualization\n",
    "fig, ax = plt.subplots(figsize=(8, 6))\n",
    "ax.plot(cat_gbm_fpr, cat_gbm_tpr, label='AUC: ' + str(cat_gbm_auc))\n",
    "ax.plot(cat_gbm_fpr, cat_gbm_fpr, 'k:')\n",
    "ax.set_xlabel('False Positive Rate', fontsize=12)\n",
    "ax.set_ylabel('True Positive Rate', fontsize=12)\n",
    "ax.legend(fontsize=12)\n",
    "ax.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}