I have a set of n samples, each associated with a set of x observations (real numbers) and y probabilities of outcomes (0 to 1 for each). I want to find subsets of observations that predict particular outcomes, where the number of subsets that predict a particular outcome is unknown.
For example, an example desired answer would be 'high values of these predictors is associated with high probability of this particular outcome.'
This seems to be some sort of mixture model clustering problem, but I'm not terribly familiar with machine learning algorithms, so thought I would see what this community thinks. This may be a question for stackoverflow instead, I'm not sure.