I am working on a project where I encountered a component which takes 96 arguments (all integer values) and outputs 12 float values. I would like to find a useful combination of these 96 values to receive the output that I want while avoiding random guessing, so the desired behavior would be that I provide the outcome and receive the 96 inputs to use them in my component.
Unfortunately I am not that experienced in that field. If I think about how I can implement this, my first thought was a kind of classification task, since I could build a dataset but the problem here is that I need integer values.
A second guess was a regression but would that be possible for y as an output vector? Are there some other approaches that could fit to my use case?