On the selection of k efficient paths by clustering techniques

In the multi-objective shortest path problem usually there is not just one single path that simultaneously minimises r >1 objective functions (costs); that is why the problem consists in finding the set of efficient paths of the given network. The number of such paths is often very large, while from an application point of view one is generally interested in selecting a given (small) number k >1 of efficient paths that are representative of the whole efficient path set. In this paper, we propose a clustering-based approach for selecting k efficient paths maximising their representativeness with respect to the cost vectors of all the efficient paths or w.r.t. the dissimilarity among the k selected paths. Computational results are given on random networks; a comparison with a known approach in the literature is also presented.