Fuzzy clustering and fuzzy entropy based classification model

In Pattern recognition, ensembles of classifiers are used to increase the performance and accuracy of classification systems. The creation of ensembles, selection of base classifiers and combining the decisions of the classifiers is an active research area. In this paper we propose a method of ensemble creation that is based on fuzzy clustering (Fuzzy C Mean) and fuzzy entropy; and named as Fuzzy Clustering and Fuzzy Entropy (FCFE) based classification model. With the help of FCM we obtained fuzzy membership matrix, revealing the underlying distribution and structure of the data. The Fuzzy entropy tells us about the degree of difficulty of classification of data. This information is used in sampling the training data into core sample and boundary sample. This sampling approach induces diversity in the ensemble which contributes to higher classification accuracy. The proposed method is evaluated on 4 UCI benchmark data sets with support vector machine (SVM) as the base classifier. The decision is combined using mean combiner rule. The results show that the proposed method delivers higher classification accuracy than stand alone SVM and the well known ensembles techniques of Bagging and Boosting.