Improved political optimizer for complex landscapes and engineering optimization problems

• Published in: Expert systems with applications Vol. 182; p. 115178
• Main Authors: Askari, Qamar, Younas, Irfan
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 15-11-2021; Elsevier BV
• Subjects: Algorithms; Benchmarks; Convergence; Elections; Engineering optimization; Exploitation; Exploration; Global optimization; Heuristic methods; Improved algorithm; Mechanical engineering; Meta-heuristic; Optimization; Optimization algorithm; Optimization algorithms; Political optimizer; Rank tests; Switching; Global optimization; Optimization algorithm; Meta-heuristic; Political optimizer; Improved algorithm; Engineering optimization
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2021.115178

•Improved variant of PO by making algorithmic and parametric changes is proposed.•IPO enhances exploration capability and balance of PO.•Performance evaluated on CEC-2014, CEC-BC-2017, and mechanical engineering problems.•Performance is compared with 25 meta-heuristics.
Political Optimizer (PO) is a recently proposed meta-heuristic with excellent convergence speed and exploitation capability. However, it is found that PO prematurely converges for complex problems because of not giving enough time to the exploration. In this paper, the exploration capability and balance of PO are improved by making multiple modifications to propose an Improved Political Optimizer (IPO). To improve the exploration capability, the condition of an equal number of parties and constituencies is relaxed, and switching with a random member of a random party is incorporated in the party-switching phase. Moreover, the balance between exploration and exploitation is enhanced by modifying the position-updating strategy (RPPUS) in the election campaign phase and replacing the tunable party-switching rate with a self-adaptive parameter. The exploitation is further improved by utilizing the best solution of the population in the parliamentary affairs phase. In addition to improvement in PO, this paper also highlights a correction in the party-switching phase of the original PO. The performance of IPO is evaluated using 30 CEC-2014 benchmarks, 29 CEC-BC-2017 benchmarks, and 6 mechanical engineering problems. It is shown through non-parametric statistical Wilcoxon’s rank-sum test that IPO significantly outperforms PO. Moreover, IPO is also compared with 10 of the well-cited and 14 latest optimization algorithms published in 2020. It is shown by using the Friedman mean-rank test that IPO secures the first rank for both types of benchmark functions. Moreover, the comparison of IPO with PO and a few well-known algorithms for 6 of the engineering problems shows that IPO performs better or equivalently to the compared optimization algorithms.