•  
  •  
 

Abstract

This study proposes a more effective concept of Learner Performance-based Behavior (LPB). It is a new metaheuristic algorithm based on how the university admission process is done for high school students in various departments. The adaptive crossover and mutation methods were incorporated into the LPB algorithm as part of an investigation. The goal is to enhance convergence and significantly improve the quality of the solutions. The aLPB (adaptive-learner performance-based Behaviour) method stands out because it sets the crossover and mutation parameters based on the performance of the parent solutions. Thus, the proposed technique achieves a balance between exploration and exploitation in the search space. The adaptiveness of these operators enables the algorithm to be more flexible and search the space when needed (by promoting heterogeneity). It also allows modeling the search process (by targeting local refinements) according to the course of elements or models as the search continues. Numerous standard test functions are used to evaluate the performance of the aLPB, including classical test functions (TF1-TF19) and the CEC-C06 2019 benchmark suite. The results are compared with those of the regular LPB and several prominent meta-heuristic algorithms, such as the dragonfly algorithm (DA), genetic algorithm (GA), and particle swarm optimization (PSO). Using the features of the proposed algorithm, one sees that the proposed crossover and mutation dynamics enhance the rate of convergence. They also cut down the trappings in local minima, as well as improving exploration and exploitation. The statistical tests (Sign and Wilcoxon) prove that the aLPB is very successful in enhancing optimization efficiency, particularly in the case of large-scale problems. The mean rank of aLPB was 2.68, and its sum rank was 51, which was better than other competitors in 19 benchmark functions. It also topped the top functions in 7 out of 10 of the CEC-C06 2019 functions. In practice, it managed to obtain a minimum cost of 263.8958 on the three-bar truss and to obtain 91 distinct solutions to the N-Queen problem, with N = 8 (as compared to 35 using GA). Results of real-world case studies also confirm its strength and versatility, making it an ideal approach for addressing multidimensional optimisation problems.

Download

Share

COinS