On solving the multiple p-median problem based on biclustering

Abstract

In this paper, we discuss the multiple p-median problem (MPMP), an extension of the original p-median problem and present several potential applications. The objective of the well-known p-median problem is to locate p facilities in order to minimize the total distance between demand points and facilities. Each demand point should be covered by its closest facility. In the MPMP, each demand point should be covered by more than one facilities closer to it, represented in total by the mc parameter. The MPMP can be applied to various location problems, e.g. the provision of emergency services where alternative facilities need to hedge against the unavailability of the primary facility, as well as to other domains, e.g. recommender systems where it may be desirable to respond to each user query with more than one available choice that satisfy their preferences. We efficiently solve the MPMP by using a biclustering heuristic, which creates biclusters from the distance matrix. In the proposed approach, a bicluster represents a subset of demand points covered by a subset of facilities. The heuristic selects appropriate biclusters taking into account the objective of the problem. Based on experimental tests performed in known benchmark problems, we observed that our method provides solutions slightly inferior to the optimal ones in significantly less computational time when compared to the CPLEX optimizer. In larger test instances, our method outperforms CPLEX both in terms of computational time and solution quality, when a time bound of 1 h is set for obtaining a solution.