Uniform Design-Based Interval Type-2 Neuro-fuzzy System and Its Performance Verification

摘要

In this paper, an interval type-2 neuro-fuzzy system with uniform design-based rule generation approach is proposed, and Begian-Melek-Mendel method is used for defuzzification. To the consequent learning, two least squares methods are involved in the consequent design for the interval type-2 neuro-fuzzy system, one is the recursive singular value decomposition, and the other is the weighted least squares estimator. Besides the interval type-2 neuro-fuzzy system modelling, another aim of this paper is to verify the interval type-2 neuro-fuzzy systems' performance from the point of view of statistics, not just the average prediction accuracy or the best results from hundreds of iteration. With this in mind, a distribution-free least squares estimation method is used to assess the interval type-2 neuro-fuzzy system's modelling capability.