Multi-Population Kidney-Inspired Algorithm With Migration Policy Selections for Feature Selection Problems

Optimization algorithms often encounter challenges in effectively managing the trade-off between exploration and exploitation, usually leading to less-than-optimal outcomes. This study introduces two novel migration policies in multi-population version of kidney-inspired algorithm (KA) to address this dilemma. The initial algorithm, coded as MultiPop-KA, implements a predetermined migration policy. Conversely, the second algorithm, coded as AutoMultiPop-KA, adopts an adaptive migration policy selection process that determines migration type based on the average fitness of sub-populations. By capitalizing on a multi-population framework and incorporating two migration policies, these methods aim to achieve a more refined equilibrium between exploration and exploitation, thereby augmenting the effectiveness of the KA. Experimental evaluations, conducted across 25 test functions and applied to 18 benchmark feature selection problems, demonstrate the efficacy of the proposed techniques. These results indicate that the proposed approach can significantly enhance optimization algorithms' performance and overall quality.