An improvement of neuro-fuzzy learning algorithm for tuning fuzzy rules

Abstract

Based on the fuzzy clustering method, we improved a neuro-fuzzy learning algorithm. In this improved approach, before learning fuzzy rules we extract typical data from training data by using fuzzy c-means clustering algorithm, in order to remove redundant data and resolve conflicts in data, and make them as practical training data. By these typical data, fuzzy rules can be tuned by using the neuro-fuzzy learning algorithm. Therefore, the learning time can be expected to be reduced and the fuzzy rules generated by the improved approach are reasonable and suitable for the identified system model. Moreover, the efficiency of the improved method is also shown by identifying nonlinear functions.