Multiple-Category Classification With Decision-Theoretic Rough Sets

Two stages with bayesian decision procedure are proposed to solve the multiple-category classification problems. The first stageis changing an m-category classification problem into m two-category classification problems, and forming three classes of rules with different actions and decisions by using of decision-theoretic rough sets with bayesian decision procedure. The second stage is choosing the best candidate rules in positive region by using the minimum probability error criterion with bayes decision theory. By considering the levels of tolerance for errors and the costs of actions in real decision procedure, we propose a new approach to deal with the multiple-category classification problems.