利用xgboost进行逻辑回归预测，基于xgb的原生接口

废话不多说，直接上代码

train = pd.read_csv('train_xy.csv')
traint, testst = train_test_split(train, test_size=0.4, random_state=1)#分割数据集

params = {
    'booster': 'gbtree',
    'objective': 'binary:logistic',
    'eval_metric': 'auc',
    'gamma': 0,
    'max_depth': depth,  # 4
    'lambda': 2,
    'subsample': 0.7,
    'colsample_bytree': 0.7,
    'min_child_weight': weight,  # 6
    'silent': 1,
    'eta': 0.007,
    'seed': 1000,  # 22很高
    'nthread': 4,
}

plst = list(params.items())
num_rounds = rounds  # 迭代次数 1000
target = testst.y
del testst['y']
y = traint.y
X = traint.drop(['y'], axis=1)
xgb_train = xgb.DMatrix(X, label=y)

xgb_test = xgb.DMatrix(testst)
model = xgb.train(plst, xgb_train, num_rounds)  # , watchlist)# ,early_stopping_rounds=100)
model.save_model('xgb.model')  # 用于存储训练出的模型
preds = model.predict(xgb_test, ntree_limit=model.best_ntree_limit)
preds1 = model.predict(xgb_test1, ntree_limit=model.best_ntree_limit)

preds1 = np.round(preds1, 6)
preds = pd.Series(preds)
test_auc = metrics.roc_auc_score(target, preds)
print(test_auc)