Development of new binary optimization algorithm using current meta heuristic optimizations for feature selection

Metaheuristic optimization methods have been used effectively in continuous optimization problems. However, these methods are difficult to be effective in solving high-dimensional and multi-modal binary optimization problems. For this reason, many binary optimization algorithms have been developed using meta-heuristics. In this study, Golden Sine Search Algorithm (Gold-SA) has been developed and Binary Gold Sine Search Algorithm (bGold-SA) has been proposed to be used in the solution of binary optimization problems. With the developed algorithm, Binary Particle Swarm Optimization (BPSO), Binary Gray Wolf Optimization (BGWO), Binary Dragonfly Algorithm (BDA), Binary Bat Algorithm (BBA) and Hybrid Binary Particle Swarm Optimization and Gravitational Search Algorithm (BPSOGSA) algorithms are compared. For the comparison process, 97 unconstrained comparison functions, 14 engineering design problems were used and an achievement test was applied using the non-parametric Wilcoxon's signed-rank test. The versions of the algorithms used in feature selection were used on 17 data sets to evaluate success. Experimental results show that bGold-SA has higher performance on comparison functions, engineering design problems and Wilcoxon's signed rank test compared to other binary algorithms. However, it was concluded that it was not as effective as BBA and BPSO in feature selection.