Stacking集成学习挑战天池新人赛【工业蒸汽量预测 】 (3) 初级学习模型构建

本项目暂时选取RandomForestRegressor，GradientBoostingRegressor和xgboost三种模型作为初级学习模型。

导入所需库

from sklearn.ensemble import RandomForestRegressor,  GradientBoostingRegressor
import xgboost as xgb

设置参数

# RandomForest parameters
rf_params = {
    'n_estimators': 100,
    'max_depth': 8
}
# GradientBoosting parameters
gb_params = {
    'n_estimators': 100,
    'max_depth': 8
}
# xgboost parameters
xgb_params = {
    'n_estimators': 100,
    'max_depth': 8,
    'eval_metric': 'rmse'
}

构建模型

rf_model = SklearnHelper(clf=RandomForestRegressor, seed=0, params=rf_params)
gb_model = SklearnHelper(clf=GradientBoostingRegressor, seed=0, params=gb_params)

# XGB分为原生接口和Sklearn接口，下面是基于Sklearn的接口, 方便后面GridSearchCV调参
xgb_model = SklearnHelper(clf=xgb.XGBRegressor, params=xgb_params)

通过交叉验证的方式训练并获取次级训练、测试的数据集

rf_oof_train, rf_oof_test = get_oof(rf_model, train_x, train_y, test_x)
gb_oof_train, gb_oof_test = get_oof(gb_model, train_x, train_y, test_x)
xgb_oof_train, xgb_oof_test = get_oof(xgb_model, train_x, train_y, test_x)

下一篇介绍次级学习模型的构建