EVOLVING SUPPORT VECTOR MACHINE USING MODIFIED FRUIT FLY OPTIMIZATION ALGORITHM AND GENETIC ALGORITHM FOR BINARY CLASSIFICATION PROBLEM

Binary classification problem is one of the mainstream research in pattern recognition field. This study proposed a modified fruit-fly optimization algorithm (FOA), which can find an eligible begin location of the FOA as starting location before running the FOA's procedure, and in the FOA's processing, the SVM parameters is modified by dynamically updating the location of each fruit-fly and the optimal feature subset is changed by evolutionary process of genetic algorithm (GA) at the same time. In the proposed method, a weighted objective function is designed to evaluate population and to take account the trade-off between sensitivity and specificity. The best individual is used to guide for evolutionary process of FOA and GA. To evaluate the classification performance of the proposed approach, this study designs several groups of comparative experiments using the proposed approach with the well-known methods, and the four binary classification datasets from the UCI Machine Learning data repository are used for the experiments. The empirical results show that the proposed approach achieves better results in terms of classification performance and a low computational cost on solving the binary classification problems.