A Bayesian Logistic Mixture Model for the Classification of Binary Outcomes