Heuristic Optimization-Based Clustering Solution for Large Facility Location Problems

Abstract

Facility location-allocation problems are one of the most important decision making areas in the supply chain management. Determining the location of the facilities and the assignment of customers to these facilities affect the cap of achievable profitability for most of the companies' supply chains. Geographical clustering of the customers, while considering their demands, has been proved to be an effective method for the facility location problem. Heuristic optimization algorithms employ an objective function that is provided by user, therefore when the total transportation cost is selected as the objective function, their performance on facility location problems is considered to be promising. The disadvantage of population based heuristic optimization algorithms on clustering analysis is their requirement of the increased number of dimensions to represent the complete solution in a single member of the population. Thus in two-dimensional geographical clustering, number of dimensions required for each population member is double of the number of required facility. In this study, a new neighborhood structure for the standard particle swarm optimization algorithm is presented for uncapacitated planar multiple facility location problem. This new approach obsoletes the need for higher number of dimensions in particles. Proposed method is benchmarked against k-means, fuzzy c-means, fuzzy c-means & center of gravity hybrid method, revised weighted fuzzy c-means and the standard particle swarm optimization algorithms on several large data sets from the literature. The results indicate that the proposed approach achieves lower total transportation cost within less computational time in facility location problems compared with the standard particle swarm optimization algorithm.