improvement of rule generation methods for fuzzy controller

This paper proposes a fuzzy modeling using the data obtained. the fuzzy system is known as a knowledge-based or rule-based system. the most important part of the fuzzy system is rule-base. one of the problems of generation of the fuzzy rules using the training data is the inconsistency of the data. existence of inconsistent and uncertain states in training data causes a high error in modeling. here, the probability fuzzy system is presented to improve the above-mentioned challenge. a zero-order sugeno fuzzy model is used as the fuzzy system structure. at first, using clustering, the number of rules and input membership functions is obtained. a set of candidate amounts is considered for the consequent parts of the fuzzy rules. considering each pair of training data, the probability of the consequent candidates is changed. in the next step, the eligibility probability of each consequent candidate is determined for all rules. finally, using the probabilities obtained, two probable outputs are generated for each input. the experimental results obtained show the superiority of the proposed approach over some available well-known approaches that reduce the number of rules and reduce the system complexity.