Optimal Parameters Selection of Support Vector Machines Using Bat Algorithm

The parameters selection process is a global combinatorial optimization problem which positively affects classification accuracy in many areas of science, especially in artificial intelligence and machine learning. In this paper, we propose a two-stage BA-SVM method, where the recent Bat Algorithm (BA) has been exploited to seek optimal parameters of Support Vector Machines (SVMs) in the first phase of BA-SVM, while the One-Versus-One method has been utilized in the second phase to generate acceptable classification outcomes. The presented method is spread on standard benchmarks and compared with three techniques from the literature. Experiments show that the BA-SVM approach was superior in all cases compared to the classification methods from the literature.