Building Fuzzy Classifiers with Pairwise Multiclass Support Vector Machines

In this paper, we make a proposal to build a classifier with fuzzy outputs with multiclass Sup- port Vector Machines (SVMs). This allows us to widen the applicability of this kind of powerful and soundly founded classifiers. We consider the pairwise multiclass version and investigate dier- ent alternatives for the aggregation process. A number of experiments have been carried out to establish the validity and performance of this ap- proach.