Jumping gene adaptations of NSGA-II with altruism approach: Performance comparison and application to williams-otto process

Abstract

Elitist non-dominated sorting genetic algorithm (NSGA-II) has been widely used for solving many application problems with multiple objectives. The concept of jumping gene (JG) from natural genetics has been incorporated into NSGA-II to improve its performance. Several JG adaptations have been proposed and used to solve multi-objective optimization test/application problems; aJG, saJG and sJG are recent JG adaptations, and they have similar performance. Further, the concept of altruism, inspired by the honey bee colony, has been incorporated with NSGA-II-aJG, and it has improved the search performance. In the present work, Alt-NSGA-II-aJG is modified for using saJG and sJG adaptations, and then performances of Alt-NSGA-II-aJG, Alt-NSGA-II-saJG and Alt-NSGA-II-sJG algorithms are compared on test and application problems. In the literature, the maximum number of generations (MNG) is the commonly used termination criterion for stochastic search algorithms. Hence, a search termination criterion based on the improvement in the Pareto-optimal front obtained has been included in the present study. Performance of selected algorithms is compared using both improvement-based termination criterion and MNG; here, generational distance, spread and inverse generational distance are employed to assess the quality of non-dominated solutions obtained. Results show that performance of Alt-NSGA-II-aJG, Alt-NSGA-II-saJG and Alt-NSGA-II-sJG algorithms is comparable, and use of the altruism approach and improvement-based termination criterion enhances the search algorithm significantly.