ranking averages是怎么提升模型性能的?


我现在在看一篇讲ensemble的文章:https://mlwave.com/kaggle-ensembling-guide/

有句话我不太理解:Ranking averages do well on ranking and threshold-based metrics (like AUC) and search-engine quality metrics (like average precision at k).

评估方式是ranking或者AUC的情况下,ranking averages为什么可以提升效果?是怎么做到的?
已邀请:

要回复问题请先登录注册