Advanced backtracking search for solving continuous optimization problems.

This paper recommends developing advanced backtracking search (ABS) to use single- and multi-vector mutation strategies to effectively enhance the backtracking search algorithm in solving a variety of optimization problems. The ABS version proposed in this paper utilizes three primary strategies considering critical historical information and suitable crossover mechanisms. Two of these are single-vector strategies that conduct searches based on random individuals, respectively, heading toward historical positions and determining destinations with one perturbation vector. The remaining multi-vector strategy conducts searches around historical best positions using a relatively low crossover rate. The performance of the suggested six ABS versions was evaluated using the benchmark functions of IEEE CEC2005 and CEC2019. The experimental results demonstrate that the proposed ABS version significantly improves BSA and its improved version. Additionally, the proposed ABS version is the most competitive algorithm compared to the seven classical algorithms in terms of evaluations of obtained results and significant values. [ABSTRACT FROM AUTHOR]