Learning Hierarchical Fuzzy Rule-Based Classification Systems By Evolutionary Boosting Algorithm

In this paper, a new hierarchical fuzzy rule-based classification system based on evolutionary boosting algorithm is proposed. the rules of the proposed system are created in different levels hierarchically in a way that their membership functions are in different sizes based on the level.the flexible linguistic value definition helps the proposed classifier system to generate coarse and fine fuzzy subspaces simultaneously, which causes that the problem space to be covered in a proper manner. each hierarchical fuzzy rule of the proposed system is formed by a running genetic algorithm, in which each chromosome represents a hierarchical fuzzy rule. after running the genetic algorithm, the best chromosome is chosen as a weak hypothesis for boosting algorithm. this process is repeated for other rules until all of the needed rules are learned iteratively. the performance ofhierarchical fuzzy rules generated by evolutionary boosting algorithms is evaluated by comparing the performance of the proposed algorithm and other classification methods, especially adaboost approximate classifier and hierarchical fuzzy rule based classification systems on a set of benchmarkclassification tasks. experimental results show that the proposed algorithm accomplishes high-quality results in comparison with these classification algorithms