Development of the Natural Survivor Method (NSM) for designing an updating mechanism in metaheuristic search algorithms


Kahraman H. T., Katı M., Aras S., Taşçı D. A.

ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, vol.122, 2023 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 122
  • Publication Date: 2023
  • Doi Number: 10.1016/j.engappai.2023.106121
  • Journal Name: ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Aerospace Database, Applied Science & Technology Source, Communication Abstracts, Compendex, Computer & Applied Sciences, INSPEC, Metadex, Civil Engineering Abstracts
  • Keywords: Meta-heuristic search algorithm design, Natural Survivor Method (NSM), Optimization, Real-world engineering design problems
  • Karadeniz Technical University Affiliated: Yes

Abstract

Meta-heuristic search algorithms (MHSs) are methods that take their inspiration from nature. However, the fitness value information used in the design of the update mechanism in MHSs is insufficient to represent the concept of adaptation to the environment and the ability to survive in nature. This causes problems in the selection of survivors and the premature convergence in the search process. This article introduces the Natural Survivor Method (NSM), developed as a design for population management as it occurs in nature, depending on analytical relationships and environmental factors. In the NSM, scores representing the adaptability of individuals to nature are calculated in order to determine the survivors. In this proposed method, the update mechanism is designed using NSM scores instead of fitness values. The NSM is the first study presented to the literature on this subject since the 1980s, when the meta-heuristics was introduced. The NSM was used for survivor selection by applying it to three different types of MHS algorithms based on physics (SFS), biology (TLABC), and evolution (LSHADE-SPACMA). Thirty-nine global optimization problems in the IEEE CEC 2017/2020 benchmark suites and ten constrained real-world engineering problems were used in the experimental studies. Data obtained from experimental studies were analyzed by using non-parametric statistical test methods. Among the 25 competing algorithms according to Friedman scores, the rankings of the three algorithms with NSM and their original versions are as follows: While TLABC is 18th, NSM-TLABC is 9th, SFS is 10th, NSM-SFS is 6th, LSHADE-SPACMA is 3rd, NSM-LSHADE-SPACMA is 1st. According to the results of the Wilcoxon pairwise comparison test between the original and NSM versions of the algorithms, the NSM versions have a clear advantage in finding optimal solutions. However, the drawback of the proposed method is that it increases the computational complexity of the algorithms. The source codes of the NSM (NSM-LSHADE_SPACMA, NSM-TLABC and NSM-SFS) can be accessed at this link: https://www.mathworks. com/matlabcentral/fileexchange/126050-natural-survivor-method-nsm-for-metaheuristic-algorithms.