Combining Fuzzy c-Means Classifiers Using Fuzzy Majority Vote
Haidong Yang, Chunsheng Li · 2008
Although fuzzy c-means classifier has been proved preferable to crisp ones and various types of fuzzy c-means classifiers have been designed, none of them are universal enough to perform equally well in all cases. A promising direction for more robust fuzzy c-means classification is to derive multiple candidate fuzzy c-means classification over a common dataset and then combine them into a consolidate one. This paper devotes to the combination of multiple fuzzy c-means classifiers and proposes a combination method for fuzzy classifiers based on fuzzy majority voting rule, denoted by CFCM-FMV, which is tested on several real datasets. Experimental results show that the combination of fuzzy classifiers outperforms all the participant fuzzy classifiers in some cases in terms of the majority of cluster validity indexes.