A Comparative Study Among Recursive Metaheuristics for Gene Selection

This chapter compares 4 variants of metaheuristics (RFA, EMVO, RPSO, and RBAT). The purpose is to test the impact of refinement on different types of metaheuristics (FA, MVO, PSO, and BAT). The refinement helps to enhance exploitation and to speed up the search process in multidimensional spaces. Moreover, it presents a powerful tool to solve different issues such as slow convergence. The different methods have been used for gene selection on 11 microarrays datasets to solve their various issues related to the presence of irrelevant genes. The obtained results reveal the positive impact of refinement on FA, MVO, and PSO, where all performances have been improved. On the other hand, this process harmed the BAT algorithm. The comparative study between the 4 variants highlights the efficiency of EMVO and FA in terms of precision and dimensionality reduction, respectively. Overall, this study suggests drawing attention to the choice of embedded metaheuristics in the refinement procedure, where powerful methods in exploration are recommended. Moreover, metaheuristics that risk form fast convergence are not advised.