Closed WeiFoo closed 9 years ago
what we want to see are large increases in pd without large increases in pf
i think i can see that in antv0 (for "Y") but this is a lot of data to wade thru. any succincter report?
also, some of your pds arent very good. rahul swears by smoting. maybe you should smote first (but ONLY on the training set).
use harmonic mean of pd, (1-pf), precision as the tuning goal