Improvement of Fuzzy Classification Systems Using Metaheuristic (PSO and ICA) with Dynamic Parameter Adaptation in Fuzzy Environment

In this research, fuzzy classification systems were improved by using some meta-heuristic methods with dynamic adaptation parameters. These methods were used for a set of data wherein membership functions and rules in fuzzy system are optimized by using meta-heuristic methods. In fuzzy classification method, being in different classes is assessed through membership functions which are in 0 to 1 interval. Each element belongs to the class where membership coefficient is higher. By assuming uncertainty in this research, a more realistic model for problem was made so that answers obtained from this model have more capability to be implemented. In this research, we improved fuzzy classification by using PSO1 and ICA2 methods with fuzzy parameters, which not only reduce classification error, but also provide a comparison between efficiency of fuzzy PSO and fuzzy ICA. In this research, data are not just for testing and these fuzzy algorithms were used for IRIS database (this database has 3 different classes for Iris flower and 150 data which each datum has four characteristics). We came to this conclusion that fuzzy classification systems can be improved by using fuzzy meta-heuristic methods and fuzzy ICA algorithm has more convergence velocity than fuzzy PSO algorithm