Abstract
This paper is concerned with a new architecture of an optimized FCM-based interval type-2 fuzzy neural network classifier developed with aid of Fuzzy C-Means (FCM) clustering and Particle Swarm Optimization (PSO). The premise part of the rules of this architecture is realized by two FCM clustering algorithms. These FCM clustering algorithms run for several values of the fuzzification coefficient subsequently resulting in interval type-2 membership functions. In the consequent part of the rules, the coefficients of a linear function are optimized by using a Back Propagation (BP) algorithm. The design parameters including the learning rate and the momentum term of BP as well as the fuzzification coefficients of the FCM are optimized by means of the PSO.
The proposed classifier is applied to several machine learning data, and the obtained results are compared with those produced by other classifiers reported in the literature.