abhishekkrthakur / approachingalmost

Approaching (Almost) Any Machine Learning Problem

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Why don't you use average auc (but list of auc's)?

kyxaxa opened this issue · comments

I would use the average of all aucs to exactly see what method is better, because the vector of values is more difficult to estimate.

For example, on page 136 you use feature combinations to create new features.

And it seems to you that "It seems like we have improved again". But we did not, because without that new features average auc=0.927, with new features 0.925.