回报地里，总结自己如何cracking the Data Challenge

Batistuta 发表于 2019-1-20 02:57
谢谢lz的分享，像请教一下，我看kaggle 里一般用xg boost 和light boost 的比较多，一方面有好的预测率，另 ...

可能lz只是想用h20? random forest对比gbdt在某些方面是有好处譬如说variance有可能更小，但绝大部分问题预测效果都不如gbdt。不只是比赛，在公司内的产品有用到树模型的也基本都是gbdt

太感谢了楼主！非常好的帖子

求问只是分问题用random forest 还是prediction也用random forest呢？

Great sharing.

I ran  H2OStackedEnsembleEstimator with a problem, which could be due to a bug of the H2O version I am using. Have you or anyone in the thread run H2OStackedEnsembleEstimator recently? If so, could it be run through without any problem? Thanks

感谢分享，很有用！！

谢谢楼主分享！！

Have you ever used H2OGridSearch and Stacked Ensemble?  If so, how was your experience? I liked H2OGridSearch, which did find better model with a few parameters. But could not get stacked ensemble run. Thanks

感谢分享

lz 干货满满

感谢楼主的分享