Improved starting solutions for the planar p-median problem

In this paper we present two new approaches for finding good starting solutions to the planar p-median problem. Both methods rely on a discrete approximation of continuous model that restricts facility locations given set demand points. The first method adapts phase greedy random construction algorithm proposed minimum sum squares clustering second one implements simple descent procedure based vertex exchange. resulting solution is then used as point in local search heuristic iterates between well-known Cooper?s alternating locate-allocate and transfer follow-up step with more effective selection rule. Extensive computational experiments show (1) using can significantly improve performance search, (2) hybrid combines \deep" be an strategy solving diversity problems.