Voting Algorithms Model with a Support Sets System by Class

The voting algorithms model (AlVot) allows building supervised classification methods based in partial analogies. These algorithms use a collection of features subsets as support to classify a new object, which is called support set system. Each support set consists of selected features that are intended to discriminate the class of each object in the learning matrix. In this paper, a new model called AlVot By Class (AlVot BC) is proposed. It is aimed to build a support set system by class, so that each class-specific support set provides evidence of the membership of an object to the class represented by that support set. The classification performance of the proposed algorithm is evaluated on seven databases from the UCI Machine Learning Repository. The results show a clear improvement over its analogous algorithm based on AlVot.