XGB 训练的时候添加自定义eval_metric：f1、准确率，并对样本、特征加权训练

以下demo重点说明：

• eval_metric：用以训练的时候评估，必须要指定验证集，本博文分享自定义准确率、f1
• verbose：训练的时候对验证集评估是否打印，True和1等价，比如verbose=10，就会打印n_estimators // 10次
• feature_weights：特征加权训练，给一个softmax后的特征重要性列表
• scale_pos_weight：样本加权训练，对于二分类，如果你的正负样本比为1：3，那么scale_pos_weight=3

from xgboost import XGBClassifier
from sklearn.metrics import f1_score, accuracy_score, log_loss, roc_auc_score
from sklearn.model_selection import RandomizedSearchCV

data = pd.DataFrame([{'A': 1, 'B': 2, 'C': 3, 'D': 10, 'y': 1}, {'A': 11, 'B': 22, 'C': 33, 'D': 1, 'y': 0},
                     {'A': 111, 'B': 222, 'C': 333, 'D': 1, 'y': 0}, {'A': 111, 'B': 222, 'C': 333, 'D': 1, 'y': 0},
                     {'A': 111, 'B': 21, 'C': 333, 'D': 1, 'y': 0}, {'A': 111, 'B': 90, 'C': 333, 'D': 1, 'y': 0},
                     {'A': 111, 'B': 64, 'C': 12, 'D': 1, 'y': 0}, {'A': 111, 'B': 222, 'C': 6, 'D': 1, 'y': 0}])

X, y = data.iloc[:, :-1].values, data.iloc[:, -1].values

model = XGBClassifier(booster='gbtree', objective='binary:logistic', n_jobs=4,
                      tree_method="hist", importance_type='gain',
                      n_estimators=100, learning_rate=0.01, max_depth=6,
                      gamma=0.01, reg_alpha=0.01, reg_lambda=0.01,
                      min_child_weight=20, subsample=0.9, colsample_bytree=0.9,
                      # scale_pos_weight=3,
                      )

def custom_eval(preds, dtrain):
    pred_score = 1.0 / (1.0 + np.exp(-preds))
    pred = [1 if p > 0.5 else 0 for p in pred_score]
    labels = dtrain.get_label()
    auc = roc_auc_score(labels, pred_score)
    acc = accuracy_score(labels, pred)
    f1 = f1_score(labels, pred)
    return [('accuracy', acc), ('auc', auc), ('f1', f1)]

model.fit(X, y,
          eval_set=[(X, y)], early_stopping_rounds=10,
          # eval_metric=['auc', 'logloss'],
          eval_metric=custom_eval,
          # feature_weights=[0.1, 0.2, 0.6, 0.1],
          verbose=10)

print(model.predict_proba(X))
print(model.feature_importances_)
print(model.evals_result())

随机搜索

def random_search_param(train_X, train_y):
    fixed_params = {
        'objective': 'binary:logistic',
        'booster': 'gbtree',
        'n_jobs': 4,
        'scale_pos_weight': 3,
        'use_label_encoder': False,
    }

    params = {
        'tree_method': ['hist', 'exact'],
        'importance_type': ['gain', 'weight', 'cover', 'total_gain', 'total_cover'],
        'n_estimators': [100, 200, 300, 350, 400, 500],
        'learning_rate': [0.001, 0.01, 0.015, 0.025, 0.05, 0.1, 0.2],
        'max_depth': [3, 5, 6, 7, 9, 12, 15, 17, 25],
        'gamma': [0, 0.01, 0.05, 0.1, 0,2],
        'reg_alpha': [0, 0.01, 0.05, 0.1, 0,2],
        'reg_lambda': [0, 0.01, 0.05, 0.1, 0,2],
        'min_child_weight': [3, 10, 15, 20, 30, 50, 60],
        'subsample': [0.8, 0.9, 1],
        'colsample_bytree': [0.8, 0.9, 1],
    }

    model = XGBClassifier(**fixed_params)
    optimized_GBM = RandomizedSearchCV(model, params, n_iter=50, cv=5, n_jobs=4)
    optimized_GBM.fit(train_X, train_y)

    print('参数的最佳取值：{0}'.format(optimized_GBM.best_params_))
    print('最佳模型得分:{0}'.format(optimized_GBM.best_score_))