Feature selection using hybrid Taguchi genetic algorithm and Fuzzy Support Vector Machine

The noises or the outliers in data points and the irrelevent or redundant features often reduce classification accuracy. This paper presents a novel approach of hybridizing two conventional machine learning algorithms for feature selection. Taguchi genetic algorithm (TGA) and Fuzzy Support Vector Machine (FSVM) with a new fuzzy membership function were combined in this hybrid method. The Taguchi method is an experimental design method, which is inserted between the crossover and the mutation operations of a GA to enhance the genetic algorithm so that better potential offspring can be generated. The TGA searches for the best feature set using principles of evolutionary process, after which these optimal features are then passed to the FSVM to calculate classification accuracy. Experimental results show that the presented approach is able to produce good performance on reducing the effects of the outliers and the noises and significantly improves the classification accuracy.