Closed huzza closed 5 years ago
Thanks Zhanghao for this PR :)
Could you summarize changes in this PR? Saw 18 files changed and not clear how i can review it.
There are mainly two changes - 1) set model score name according the model spec file name (before we always use model0, model1, ...); 2) for example if the positive rate is too now, user can adjust positive rate when shuffling norm/clean dataset. We achieve this by duplicate positive records or adjust the weight of positive records.