XGBoost API介绍

import xgboost as xgb

import numpy as np
import xgboost as xgb
from sklearn.datasets import make_classification, make_regression
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, mean_squared_error

# 1. sklearn风格的XGBoost
print("=== sklearn风格的XGBoost ===")

# 生成分类数据
X, y = make_classification(
    n_samples=1000,
    n_features=20,
    n_informative=10,
    n_classes=2,
    random_state=42
)

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# sklearn风格
xgb_clf = xgb.XGBClassifier(
    n_estimators=100,
    max_depth=3,
    learning_rate=0.1,
    subsample=0.8,
    colsample_bytree=0.8,
    random_state=42
)

xgb_clf.fit(X_train, y_train)
y_pred = xgb_clf.predict(X_test)
acc = accuracy_score(y_test, y_pred)
print(f"XGBoost分类准确率: {acc:.4f}")

# 2. 非sklearn风格（原生风格）
print("\n=== 非sklearn风格（原生风格） ===")

# 转换为DMatrix格式
dtrain = xgb.DMatrix(X_train, label=y_train)
dtest = xgb.DMatrix(X_test, label=y_test)

# 参数设置
params = {
    'max_depth': 3,
    'eta': 0.1,  # 学习率
    'objective': 'binary:logistic',
    'eval_metric': 'logloss',
    'subsample': 0.8,
    'colsample_bytree': 0.8
}

# 训练
num_round = 100
bst = xgb.train(params, dtrain, num_round)

# 预测
y_pred_proba = bst.predict(dtest)
y_pred = (y_pred_proba > 0.5).astype(int)
acc = accuracy_score(y_test, y_pred)
print(f"XGBoost原生风格准确率: {acc:.4f}")

# 3. XGBoost回归
print("\n=== XGBoost回归 ===")

X, y = make_regression(
    n_samples=1000,
    n_features=20,
    n_informative=10,
    noise=0.1,
    random_state=42
)

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# sklearn风格
xgb_reg = xgb.XGBRegressor(
    n_estimators=100,
    max_depth=3,
    learning_rate=0.1,
    random_state=42
)

xgb_reg.fit(X_train, y_train)
y_pred = xgb_reg.predict(X_test)
mse = mean_squared_error(y_test, y_pred)
print(f"XGBoost回归MSE: {mse:.4f}")

# 4. 多分类
print("\n=== XGBoost多分类 ===")

X, y = make_classification(
    n_samples=1000,
    n_features=20,
    n_informative=10,
    n_classes=3,
    n_clusters_per_class=1,
    random_state=42
)

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# 多分类
xgb_multi = xgb.XGBClassifier(
    n_estimators=100,
    max_depth=3,
    learning_rate=0.1,
    objective='multi:softprob',
    num_class=3,
    random_state=42
)

xgb_multi.fit(X_train, y_train)
y_pred = xgb_multi.predict(X_test)
acc = accuracy_score(y_test, y_pred)
print(f"XGBoost多分类准确率: {acc:.4f}")

# 5. 参数对比
print("\n=== 参数对比 ===")
print("sklearn风格 vs 原生风格:")
print("  n_estimators  vs  num_boost_round")
print("  max_depth     vs  max_depth")
print("  learning_rate vs  eta")
print("  subsample     vs  subsample")
print("  colsample_bytree vs colsample_bytree")

# 6. 模型保存和加载
print("\n=== 模型保存和加载 ===")

# 保存
xgb_clf.save_model('xgboost_model.json')
print("模型已保存")

# 加载
xgb_loaded = xgb.XGBClassifier()
xgb_loaded.load_model('xgboost_model.json')
y_pred = xgb_loaded.predict(X_test)
acc = accuracy_score(y_test, y_pred)
print(f"加载后准确率: {acc:.4f}")

# 7. 特征重要性
print("\n=== 特征重要性 ===")
importance = xgb_clf.feature_importances_
print("特征重要性（前5个）:")
for i in np.argsort(importance)[-5:][::-1]:
    print(f"  特征{i}: {importance[i]:.4f}")

# 8. 早停
print("\n=== 早停 ===")
X, y = make_classification(
    n_samples=1000,
    n_features=20,
    n_informative=10,
    n_classes=2,
    random_state=42
)

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42
)

# 进一步划分验证集
X_train, X_val, y_train, y_val = train_test_split(
    X_train, y_train, test_size=0.2, random_state=42
)

xgb_es = xgb.XGBClassifier(
    n_estimators=1000,
    max_depth=3,
    learning_rate=0.1,
    early_stopping_rounds=10,
    eval_metric='logloss',
    random_state=42
)

xgb_es.fit(
    X_train, y_train,
    eval_set=[(X_val, y_val)],
    verbose=False
)

y_pred = xgb_es.predict(X_test)
acc = accuracy_score(y_test, y_pred)
print(f"早停后准确率: {acc:.4f}")
print(f"最佳轮数: {xgb_es.best_iteration}")