rov-offline/ad_xgb_model_params_search.py

import datetime
import pandas as pd
from xgboost.sklearn import XGBClassifier
from sklearn.model_selection import GridSearchCV
from sklearn import metrics


now_date = datetime.datetime.today()
dt = datetime.datetime.strftime(now_date, '%Y%m%d')
# 1. 读取数据
train_data = pd.read_csv(f'./data/train_test_data/train_{dt}.csv')
print(train_data.shape)
test_data = pd.read_csv(f'./data/train_test_data/test_{dt}.csv')
print(test_data.shape)
# 2. 划分x和y
data_columns = train_data.columns.values.tolist()
x_train = train_data[data_columns[:-1]]
y_train = train_data[data_columns[-1]]
x_test = test_data[data_columns[:-1]]
y_test = test_data[data_columns[-1]]
print(f"x_train_shape: {x_train.shape}")
print(f"x_test_shape: {x_test.shape}")

parameters = {
    'max_depth': range(3, 10, 2),  # 树的最大深度
    'learning_rate': [i/10 for i in range(1, 5)],  # 学习率
    'n_estimators': range(50, 500, 50),  # 最大迭代次数
    'min_child_weight': range(1, 6, 2),  # 叶子结点的最小权重
}

grid = GridSearchCV(
    estimator=XGBClassifier(
        learning_rate=0.1,  # 学习率，控制每次迭代更新权重时的步长，默认0.3;调参：值越小，训练越慢;典型值为0.01-0.2
        n_estimators=50,  # 总共迭代的次数，即决策树的个数
        max_depth=3,  # 树的深度,默认值为6，典型值3-10;调参：值越大，越容易过拟合；值越小，越容易欠拟合
        min_child_weight=5,  # 叶子节点最小权重;默认值为1;调参：值越大，越容易欠拟合；值越小，越容易过拟合
        objective='binary:logistic',
        eval_metric=['error', 'auc']
    ),
    param_grid=parameters,
    scoring='roc_auc',
    n_jobs=4,
    iid=False,
    cv=5)
grid.fit(x_train, y_train)

print(f"grid.best_params_: {grid.best_params_}")
print(f"grid.best_score_: {grid.best_score_}")

clf = grid.best_estimator_
print(f"grid.best_estimator_: {grid.best_estimator_}")

clf.fit(x_train, y_train)
y_train_pred = clf.predict(x_train)
train_accuracy = metrics.accuracy_score(y_train, y_train_pred)
train_auc = metrics.roc_auc_score(y_train, y_train_pred)
train_recall = metrics.recall_score(y_train, y_train_pred)
train_f1 = metrics.f1_score(y_train, y_train_pred)
train_precision = metrics.precision_score(y_train, y_train_pred)
print(f"Train Accuracy: {train_accuracy * 100.0}.2f%%, "
      f"auc: {train_auc}.2f, "
      f"recall: {train_recall * 100.0}.2f%%, "
      f"f1: {train_f1 * 100.0}.2f%%, "
      f"precision: {train_precision * 100.0}.2f%%")

y_test_pred = clf.predict(x_test)
test_accuracy = metrics.accuracy_score(y_test, y_test_pred)
test_auc = metrics.roc_auc_score(y_test, y_test_pred)
test_recall = metrics.recall_score(y_test, y_test_pred)
test_f1 = metrics.f1_score(y_test, y_test_pred)
test_precision = metrics.precision_score(y_test, y_test_pred)
print(f"Test Accuracy: {test_accuracy * 100.0}.2f%%, "
      f"auc: {test_auc}.2f, "
      f"recall: {test_recall * 100.0}.2f%%, "
      f"f1: {test_f1 * 100.0}.2f%%, "
      f"precision: {test_precision * 100.0}.2f%%")