Solving a continuous location–routing problem by use of a self-organizing map

Purpose – A variant of Kohonen's algorithm for the self-organizing map (SOM) is used to solve a continuous location–routing problem that can be applied to identify potential sites for subsequent selection by a discrete finite set model. The paper aims to show how the algorithm may be customized to fit the problem structure in a way that allows aspects of location and routing to be integrated into the solution procedure. Design/methodology/approach – A set of test instances is used to compare the solutions of the neural network to those obtained by sequential approaches based on a savings procedure. Findings – Compared to the results of the sequential approaches, the neural network yields good results. Research limitations/implications – Future work may cover the expansion of the neural approach to multi-depot and multi-stage problems. Additionally, application of procedures other than the savings procedure should be evaluated with respect to their potential for further enhancing the solution quality of the sequential approaches. Practical implications – This paper shows that strategic location decisions in practical applications with long-term customer relationships can be taken using simultaneously generated routing information on an operational level. Originality/value – The paper provides a new variety of applications for SOM as well as high quality results for the specific type of problem considered.