Learning Rough Set Based Classifiers Using Boolean Kernels

In this paper we present a hybridization of Rough Set (RS) theory and Support Vector Machine (SVM). Both approaches to data analysis employ the area between positively and negatively labeled examples, i.e. the "boundary region" in RS and the "margin" in SVM, but they offer different ways to use this concept in the classification problem. We will show that despite differences, many Rough Set methods can be also implemented by SVM. In particular we will show that the rough set methodology to discretization problem can be also solved by SVM with a special Boolean kernel. At the end we propose a compound classification method that aggregates the feature selection method in RS and object selection method in SVM.