Evolving the SVM model based on a hybrid method using swarm optimization techniques in combination with a genetic algorithm for medical diagnosis

Abstract

In this paper, we introduce a novel hybrid method that uses a genetic algorithm (GA) in combination with a swarm optimization algorithm (particle swarm optimization (PSO) or fruit fly optimization algorithm (FOA)) for medical diagnosis. The proposed approaches, called GAPSO-FS and GAFOA-FS, simultaneously employ the genetic algorithm (GA) to choose the optimal feature subset and the swarm optimization algorithms (PSO/FOA) to optimize the SVM parameters. This procedure primarily comprises three synchronized parallel layers, including two optimization layers and an intermediate layer. The intermediate layer is mainly responsible for harmonizing the information from the two optimization layers and then distributing the processed information back to those layers. The major contribution of the proposed approaches is that they fully exploit the advantages of the different algorithms. The genetic algorithm (GA) excels at selecting the optimal feature subset, whereas swarm optimization algorithms (PSO/FOA) are optimal for searching the most appropriate continuous variables, including the penalty parameter C and the hyperplane parameter. We performed several groups of experiments on real-world medical cases from the UCI machine learning data repository to compare our hybrid approaches with well-known optimization techniques. The empirical results demonstrated that the proposed GAPSO-FS and GAFOA-FS can select the best SVM model parameters and a more highly relevant feature subset for the SVM classifier than a single algorithm can, thus improving the classification performance when solving a medical diagnosis problem. Therefore, the proposed approach has potential as a useful tool in medical diagnosis.