mLPB: A Modified Learner Performance-Based Behavior Algorithm Through Differential Evolution

Abstract

This research work presents a modified learner performance-based behavior (mLPB). It blends the learner performance-based behavior (LPB) method with the differential evolution (DE) algorithm to achieve better optimization performance. Although the LPB algorithm performs well, its limitations in the exploitation phase and slower convergence prevent it from achieving optimal results. These issues must be addressed, as they render the LPB algorithm less competitive compared to other well-established algorithms designed to tackle complex optimization problems. Inspired by the aforementioned LPB algorithm, but aiming to enhance its performance further, we replace its crossover and mutation operations with those of the DE to foster a better equilibrium between the phases of exploration and exploitation. Extensive experiments are conducted to evaluate the performance of mLPB in comparison to Genetic Algorithm (GA), Dragonfly Algorithm (DA), and Particle Swarm Optimization (PSO) on various benchmark functions, including the CEC-C06 2019 test functions. In addition, the study reviews other leading methods, including FDO, GOOSE, Leo, IFDO, FOX, CLPB, WOA, and SSA. Following this, the method works on engineering problems such as designing pressure vessels and solving the traveling salesman problem (TSP). It is quite clear from the findings that the algorithm we tested works better than the others tried in the study. Based on Wilcoxon tests, mLPB increases the processing speed, speed of convergence, and all aspects of optimization, outperforming most of the algorithms in solving large problems.