Developing parsimonious ensembles using predictor diversity within a reinforcement learning framework