{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"- **Author：** 马肖\n",
"- **E-Mail：** maxiaoscut@aliyun.com\n",
"- **GitHub：** https://github.com/Albertsr"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np\n",
"import pandas as pd\n",
"from scipy.sparse import hstack\n",
"from sklearn.datasets import make_classification\n",
"from sklearn.ensemble.gradient_boosting import GradientBoostingClassifier\n",
"from sklearn.linear_model import LogisticRegression\n",
"from sklearn.metrics import roc_auc_score, accuracy_score, f1_score\n",
"from sklearn.model_selection import train_test_split\n",
"from sklearn.preprocessing import OneHotEncoder\n",
"from xgboost import XGBClassifier\n",
"from lightgbm import LGBMClassifier\n",
"\n",
"X, y = make_classification(n_samples=10000, n_features=20, n_informative=18, n_redundant=2,\n",
" n_classes=2, n_clusters_per_class=3, random_state=2017)\n",
"X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25, random_state=0)"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [],
"source": [
"clf_gbdt = GradientBoostingClassifier(n_estimators=50)\n",
"clf_xgb = XGBClassifier(n_estimators=50)\n",
"clf_lgb = LGBMClassifier(n_estimators=50)\n",
"lr = LogisticRegression(max_iter=500, solver='lbfgs')\n",
"\n",
"models = [clf_gbdt, clf_xgb, clf_lgb]\n",
"names = ['GBDT', 'XGBoost', 'LightGBM']\n",
"\n",
"metric_scores = []\n",
"for model,name in zip(models, names):\n",
" model.fit(X_train, y_train)\n",
" y_pred = model.predict(X_test)\n",
" y_pred_prob = model.predict_proba(X_test)[:, 1]\n",
" acc = accuracy_score(y_test, y_pred)\n",
" auc = roc_auc_score(y_test, y_pred_prob)\n",
" fscore = f1_score(y_test, y_pred)\n",
"\n",
" if name == 'GBDT':\n",
" X_train_leaves = model.apply(X_train)[:, :, 0]\n",
" X_test_leaves = model.apply(X_test)[:, :, 0]\n",
" \n",
" elif name == 'LightGBM':\n",
" X_train_leaves = model.predict(X_train, pred_leaf=True)\n",
" X_test_leaves = model.predict(X_test, pred_leaf=True)\n",
" else:\n",
" X_train_leaves = model.apply(X_train)\n",
" X_test_leaves = model.apply(X_test)\n",
"\n",
" \n",
" All_leaves = np.r_[X_train_leaves, X_test_leaves]\n",
" All_leaves = All_leaves.astype(np.int32)\n",
"\n",
" enc = OneHotEncoder(categories='auto')\n",
" X_new_feat = enc.fit_transform(All_leaves)\n",
" \n",
" train_samples = X_train_leaves.shape[0]\n",
" X_train_new = X_new_feat[:train_samples, :]\n",
" X_test_new = X_new_feat[train_samples:, :]\n",
"\n",
" X_train_hstack = hstack([X_train_new, X_train])\n",
" X_test_hstack = hstack([X_test_new, X_test])\n",
"\n",
" lr.fit(X_train_hstack, y_train)\n",
" y_pred_2 = lr.predict(X_test_hstack)\n",
" y_pred_prob_2 = lr.predict_proba(X_test_hstack)[:, 1]\n",
"\n",
" new_acc = accuracy_score(y_test, y_pred_2)\n",
" new_auc = roc_auc_score(y_test, y_pred_prob_2)\n",
" new_fscore = f1_score(y_test, y_pred_2)\n",
" score = {'OriginalFeature':[fscore, acc, auc], 'NewFeature':[ new_fscore, new_acc, new_auc]}\n",
" result = pd.DataFrame(score)\n",
" metric_scores.append(result)"
]
},
{
"cell_type": "code",
"execution_count": 3,
"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></th>\n",
" <th>OriginalFeature</th>\n",
" <th>NewFeature</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th rowspan=\"3\" valign=\"top\">GBDT + LR</th>\n",
" <th>F1</th>\n",
" <td>0.841070</td>\n",
" <td>0.875536</td>\n",
" </tr>\n",
" <tr>\n",
" <th>ACC</th>\n",
" <td>0.838400</td>\n",
" <td>0.872400</td>\n",
" </tr>\n",
" <tr>\n",
" <th>AUC</th>\n",
" <td>0.925139</td>\n",
" <td>0.946116</td>\n",
" </tr>\n",
" <tr>\n",
" <th rowspan=\"3\" valign=\"top\">XGBoost + LR</th>\n",
" <th>F1</th>\n",
" <td>0.837136</td>\n",
" <td>0.872116</td>\n",
" </tr>\n",
" <tr>\n",
" <th>ACC</th>\n",
" <td>0.834400</td>\n",
" <td>0.869200</td>\n",
" </tr>\n",
" <tr>\n",
" <th>AUC</th>\n",
" <td>0.921574</td>\n",
" <td>0.943909</td>\n",
" </tr>\n",
" <tr>\n",
" <th rowspan=\"3\" valign=\"top\">LightGBM + LR</th>\n",
" <th>F1</th>\n",
" <td>0.910658</td>\n",
" <td>0.921269</td>\n",
" </tr>\n",
" <tr>\n",
" <th>ACC</th>\n",
" <td>0.908800</td>\n",
" <td>0.919600</td>\n",
" </tr>\n",
" <tr>\n",
" <th>AUC</th>\n",
" <td>0.969011</td>\n",
" <td>0.971790</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" OriginalFeature NewFeature\n",
"GBDT + LR F1 0.841070 0.875536\n",
" ACC 0.838400 0.872400\n",
" AUC 0.925139 0.946116\n",
"XGBoost + LR F1 0.837136 0.872116\n",
" ACC 0.834400 0.869200\n",
" AUC 0.921574 0.943909\n",
"LightGBM + LR F1 0.910658 0.921269\n",
" ACC 0.908800 0.919600\n",
" AUC 0.969011 0.971790"
]
},
"execution_count": 3,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"model_names = ['GBDT + LR', 'XGBoost + LR', 'LightGBM + LR'] \n",
"model_metrics = ['F1', 'ACC', 'AUC']\n",
"col_idx = pd.MultiIndex.from_product([model_names, model_metrics])\n",
"df_contrast = pd.concat(metric_scores, axis=0)\n",
"df_contrast.index = col_idx \n",
"df_contrast"
]
}
],
"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": "3.7.1"
}
},
"nbformat": 4,
"nbformat_minor": 2
}