Fuzzy Combination Rule of Multiple Classifiers System Based on Yager Triangular Norm

This article focuses on the fuzzy combination rule of multiple classification system. As one of the fuzzy triangular norm operation model, Yager triangular norm can improve the generalization capability of pattern classification systems. We extend the Yager t-norm from binary form to multivariate weighted form. Based on it, we put forward a new fuzzy combination rule. The new rule is a set of continuously changing operator clusters that can make better use of the classified information of sub-classifiers. It generates different computing models by the different characteristics of datasets to combine sub-classifiers. We use genetic algorithm to evaluate the parameters of the new rule and use UCI standard datasets to test it. The experimental results show that our rule leads to less error and better performance than product rule, mean rule, median rule, and majority vote rule.