Features Selection using Fuzzy ESVDF for Data Dimensionality Reduction

This paper introduces a novel algorithm for features selection based on a Support Vector Decision Function (SVDF) and Forward Selection (FS) approach with a fuzzy inferencing model. In the new algorithm, Fuzzy Enhancing Support Vector Decision Function (Fuzzy ESVDF), features are selected stepwise, one at a time, by using SVDF to evaluate the weight value of each specified candidate feature, then applying FS with the fuzzy inferencing model to rank the feature according to a set of rules based on a comparison of performance. Using a fast and simple approach, the Fuzzy ESVDF algorithm produces an efficient features set and, thus, provides an effective solution to the dimensionality reduction problem in general. We have examined the feasibility of our approach by conducting several experiments using five different datasets. The experimental results indicate that the proposed algorithm can deliver a satisfactory performance in terms of classification accuracy, False Positive Rate (FPR), training time, and testing time.