Membership Function Modification in Fuzzy SVM using Combination of Distance Feature, Correlation, and Depth of Data

In this paper, we proposed the membership function modification in Fuzzy Support Vector Machine (Fuzzy SVM) based on exponential distribution function. This function is expressed as the level of training data contribution to its class model building. It is formed prior to training phase and started with checking and handling the missing values. Afterwards is the formation of fuzzy variables based on the distance feature, correlation, and the depth of data to be formalized into the exponential distribution function. The experiments results show that the proposed method has a higher accuracy about 2% compared to conventional SVM and Fuzzy SVM. This method is expected more contributed in the development of various pattern recognition applications, which more demand on higher accuracy. Index Term-Fuzzy, svm, feature combination.