Interpretation of Support Vector Machines by means of Fuzzy Rule-Based Systems
Juan Luis Castro, L. D. Flores-Hidalgo, Carlos Javier Mantas · European Society for Fuzzy Logic and Technology Conference · 2005
Support Vector Machines (SVM) have demonstrated their ability in solving classiflcation problems in an optimal way with a solid mathematical background. In this paper we improve the interpretability of SVM’s by showing that every SVM is exactly represented by a Fuzzy Rule BasedSystem, for every kernel function used. Nevertheless, this system is in some way compact in their rules and for that reason, we introduce another FRBS, called ‐-FRBS, that approximates it and which is suitable to decompose its rules in simple fuzzy propositions. We show it with an example in the last section.