An Improved Genetic Algorithm with a New Initialization Mechanism Based on Regression Techniques

摘要: Genetic algorithm (GA) is one of the well-known techniques from area evolutionary computation that plays a significant role in obtaining meaningful solutions to complex problems with large search space. GAs involve three fundamental operations after creating an initial population, namely selection, crossover, and mutation. The first task create appropriate population. Traditionally randomly selected population widely used as it simple efficient; however, generated may contain poor fitness. Low quality or fitness individuals lead take long time converge optimal (or near-optimal) solution. Therefore, determining near-optimal In this work, we propose new method for seeding based on linear regression analysis problem tackled by GA; paper, traveling salesman (TSP). proposed Regression-based technique divides given scale TSP into smaller sub-problems. This done using line its perpendicular line, which allow clustering cities four sub-problems repeatedly, location each city determines category/cluster belongs to, works repeatedly until size subproblem becomes very small, less instance, these are more likely neighboring other, so connecting them other creates somehow good solution start with, mutated several times form We analyze performance GA when traditional techniques, such random nearest neighbors, along regression-based technique. experiments carried out some instances obtained TSPLIB, standard library problems. Quantitative statistical test tools: variance (ANOVA), Duncan multiple range (DMRT), least difference (LSD). experimental results show uses outperforms neighbor terms error rate, average convergence.