A hybrid design of shadowed type-2 fuzzy inference systems applied in diagnosis problems

Abstract Computer aided systems have been frequently used in recent times and the evolution of artificial intelligence allows the inclusion of this kind of systems in most complex problems, for example in diagnosis problems. Some of the most powerful algorithms or methods that have been applied to diagnosis problems are: Artificial Neural Networks, Support Vector Machines, Decision Threes, Fuzzy Inference Systems and hybrids of these algorithms. The present paper explores the applications of a special approximation of the Type-2 Fuzzy Inference System, which is the Shadowed Type-2 Fuzzy Inference System, and the reason for using this approach (and not another one) is because it provides a good approximation to General Type-2 Fuzzy Inference Systems, but with a computational cost reduction. Shadowed Type-2 Fuzzy Inference Systems are inspired on the Shadows Sets that simplify the traditional Type-1 Membership Function using two optimal α -cuts, which in our case are applied in the secondary membership function of the General Type-2 Fuzzy Inference System (modeled by the α -planes representation) in order to model the system with only two α -planes. Experiments were realized with eleven benchmark datasets in order to evaluate the accuracy of the proposed system. The obtained results demonstrate the advantages of using this approach over the conventional General Type-2 Fuzzy Inference Systems, and this is because we obtain better performance in most of the cases and with less computational resources.