I have two sets of continuous variables on which I would like to perform a non-linear weighted combination that best predicts a binary outcome. What could be the recommended method to use and why?
Thought I made it clear, or maybe not. Ok, let's try again. I have two floating point vectors each of which is a predictor of a binary outcome (say presence or absence of a disease). My question is: Is there any chance that one can non-linearly combine the two to yield an even stronger predictor ( in other words, to fuse the information present in the two descriptors)? I hope that makes sense.