Enhancing Salp Swarm Algorithm Based on Elite Opposition based Learning and Cauchy Mutation Strategies

Abstract

The significance of machine learning models for real life scenarios to design prediction model is outstanding. However, the nature of datasets available especially in biomedical domain calls for optimization strategies for effective and reliable solution. In view of this development, so many optimization algorithms are promising but not without limitations. For instance, the salp swarm algorithm (SSA), which is a higher level procedure algorithm, has demonstrated a lot of capacity to solve optimization challenges for prediction models. However, to solve an intricate optimization issues, the SSA is deficient with the dawdling convergence pace and a tendency of diminishing into sub-optimal results. Therefore, this research aims to enhance the algorithm using Cauchy mutation and elite opposition based learning strategies. The objective is achieved by fusing these strategies into the existing algorithm; thereby resulting to an enhanced SSA optimization algorithm christened EOnCaSSA in this paper. The search agents of 50, 100, and 150 were chosen for each bench functions and 300, 600 and 900 numbers of iterations were also chosen for each function. It is noteworthy that all the seven benchmark functions for unimodal functions have an optimum value of zero. The four evaluation metrics, which include the best, worst, average and standard deviation to determine and compare the performance of EOnCaSSA presents significant improvement over the tradition SSA. This implies that the enhanced algorithm can be employed to solve both simple and difficult optimization issues as it passes all the benchmark functions tests.