A Clustering-based Simulated Annealing Algorithm with Taguchi Method for the Discrete Ordered Median Problem

Authors : Mustafa Serdar TOKSOY

Pages : 169-184

Doi:10.16984/saufenbilder.1034945

View : 16 | Download : 3

Publication Date : 2022-02-28

Article Type : Research Paper

Abstract :Researchers have studied discrete location problems for a long time because of their importance in practice. The Discrete Ordered Median Problem insert ignore into journalissuearticles values(DOMP); generalizes discrete facility location problems. The DOMP generalizes the main facility location problems` objective functions such as the p-median, p-center and p-centdian location problems. As these problems, also known as the problems of location-allocation, have NP-hard structure, it is inevitable to use heuristic methods for solution. In this study, a metaheuristic algorithmic suggestion will be put forward by examining the DOMP to find optimal solutions. For that purpose, we proposed a Simulated Annealing insert ignore into journalissuearticles values(SA); metaheuristic with K-means Clustering Algorithm in initialization for the DOMP. Novel approaches for initial solution and K-exchange algorithm-based neighborhoods for local search were analysed. In addition, best level of selected parameters were determined by Taguchi method. Forty common p-median instances derived from OR-LIB were used to test the SA performance, and the results were compared with three state-of-art algorithms in the literature. According to the computational results, 21 best solutions were obtained on instances despite gap values and CPU times increasing proportionally to the scale of the instances. In a conclusion, the proposed clustering-based SA algorithm is competitive and can be a robust alternative for the DOMP.
Keywords : Discrete ordered median problem, Simulated annealing, Taguchi, K means clustering