MODIFIED GRAVITATIONAL SEARCH ALGORITHM WITH PARTICLE MEMORY ABILITY AND ITS APPLICATION

Gravitational search algorithm (GSA) is a type of optimization algorithm based on the law of gravity and mass interactions, which is lacking of memory ability. To enhance particle memory ability and search accuracy of GSA, a modified GSA (MGSA) is developed. MGSA adopts the idea of local optimum solution and global optimum solution from particle swarm optimization (PSO) algorithm into GSA. Furthermore, the convergence property of MGSA is analyzed. The performance of MGSA has been evaluated on 12 standard benchmark functions, and the results were compared with GSA. The obtained experimental results verified the effectiveness of MGSA in solving high dimensional benchmark functions. Additionally, to test MGSA performance in practical issue, MGSA is applied into support vector machine (SVM) parameter settings, the results showed that suitable SVM parameters could be effectively found by MGSA.