An Improved Pathfinder Algorithm (ASDR-PFA) based on Adaptation of Search Dimensional Ratio for solving constrained optimization problems and optimal Feature Selection

Abstract

Abstract The pathfinder algorithm (PFA) is a recently introduced meta-heuristic approach that is mathematically modelled by the cooperative behaviour of animal groups during a search for the best food zone. The PFA procedure comprises two phases: the pathfinder phase and the follower phase. In the former phase, the pathfinder explores new regions in the search space with its versatile explorative power. And during the later phase, followers change position following the leader and their perception, as a result, it makes it easy for the algorithm to fall in local optima leading to slow convergence. To alleviate such issues, this article introduces an improved approach to PFA named ASDR-PFA with the incorporation of a parameter termed search dimensional ratio (SDR) to generate new candidate solutions using the previous best one. The power of ASDR-PFA lies in its technique of updating the SDR parameter dynamically that attunes the balance between exploration and mining ability leading to a faster convergence towards the optimum. The proficiency of the ASDR-PFA has been examined and established using a set of 16 IEEE basic benchmark functions, applied to solve six constrained optimization complications and optimal feature selection (OFS) problems as well. Furthermore, a comparative analysis is performed on the experimental results attained by the proposed approach with five contemporary meta-heuristic methods to demonstrate its superiority.