A Multistart Alternating Tabu Search for Commercial Districting
In this paper we address a class of commercial districting problems that arises in the context of the distribution of goods. The problem aims at partitioning an area of distribution, which is modeled as an embedded planar graph, into connected components, called districts. Districts are required to be mutually balanced with respect to node attributes, such as number of customers, expected demand, and service cost, and as geometrically-compact as possible, by minimizing their Euclidean diameters. To solve this problem, we propose a multistart algorithm that repeatedly constructs solutions greedily and improves them by two alternating tabu searches, one aiming at achieving feasibility through balancing and the other at maximizing district compactness. Computational experiments confirm the effectiveness of the different components of our method and show that it significantly outperforms the current state of the art, improving known upper bounds in almost all instances.
KeywordsDistricting Territory design Tabu search Heuristic algorithm Compactness
This research was supported by the Brazilian funding agencies CNPq (grant 420348/2016-6), FAPEMIG (grant TEC-APQ-02694-16) and by Google Research Latin America (grant 25111). We would also like to thank to support of the Fundação de Desenvolvimento Científico e Cultural (FUNDECC/UFLA).
- 11.Gliesch, A., Ritt, M., Moreira, M.C.O.: A genetic algorithm for fair land allocation. In: Genetic and Evolutionary Computation Conference, pp. 793–800. ACM Press (2017)Google Scholar
- 22.Shamos, M.I.: Computational Geometry. Ph.D. thesis (1978)Google Scholar
- 23.Har-Peled, S.: On the Expected Complexity of Random Convex Hulls, pp. 1–20, November 2011. http://arxiv.org/abs/1111.5340
- 27.Chou, C., Kimbrough, S.O., Sullivan-Fedock, J., Woodard, C.J., Murphy, F.H.: Using interactive evolutionary computation (IEC) with validated surrogate fitness functions for redistricting. In: Genetic and Evolutionary Computation Conference, pp. 1071–1078 (2012)Google Scholar