Sorry if this has been posed before.
Say that I've trained one classifier that has very high recall on the class I'm interested in, +1 (~90% recall), but low precision (~50%).
I have another classifier that has very high recall AND precision on the opposite class, -1 (~98%, 95%).
If I passed the data through the first classifier, it would identify said 90% of +1's. Then, pass them through the second classifier to determine which of those +1's are actually -1's.
It seems that the recall might only drop a little on +1, but the precision for would be significantly increased. The classes I'm talking about are generally imbalanced (10-15% +1, 85-90% -1) in the data I work with. Does that sound like a valid approach?
[link][8 comments]