Suppose I have a set of n elements. I want to have a model for how "desirable" a subset of those elements is when evaluated by a person. The training/testing data is a number of given subsets, each either accepted or rejected by the person. It's known that the user likely rejects it if there's too many inherently "bad" elements, not because of incompatibility amongst elements in the subset. What statistical model should I use?
I think the right idea is to keep an array of scores for each element, updating it with each training data, but how? Bayesian? That would be if each subset was rejected precisely because of one bad element, but this isn't necessarily the case.
Thanks for any insight!