×

A spanning tree heuristic for regional clustering. (English) Zbl 0825.62291

Summary: We consider the problem of subdividing a region into districts - formed by geographically contiguous sites - which are as homogeneous as possible with respect to a certain set of characteristics. One possible mathematical formulation is the following. Given a connected graph \(G\) in which a vector of characteristic is associated with each vertex, find a minimum inertia partition of the vertex-set of \(G\) into a prescribed number of connected clusters. We propose a heuristic approach based on the solution of a sequence of relaxed problems where \(G\) is replaced by one of its spanning trees. Numerical experiments on medium-large graphs arising from real-life applications show that our heuristic almost always yields partitions with smaller inertia than those generated by a well established ascending algorithm.

MSC:

62-XX Statistics
Full Text: DOI

References:

[1] Anderberg M.R., Cluster Analysis for Applications (1973) · Zbl 0299.62029
[2] Berge C., Graphes et Hypergraphes (1983)
[3] DOI: 10.2307/2529148 · doi:10.2307/2529148
[4] DOI: 10.1111/j.1475-2743.1987.tb00702.x · doi:10.1111/j.1475-2743.1987.tb00702.x
[5] Lawler E.L., Combinatorial Optimization: Networks and Matroids (1976) · Zbl 0413.90040
[6] Lebart L., Les Cahiers de l’Analyse des Données 3 pp 275– (1978)
[7] Mulvey, J.M. and Beck, M. P. Solving capacitated clustering problems. Technical Report EES-82-2. School of Engineering and Applied Science, Princeton University. · Zbl 0547.62039
[8] DOI: 10.1093/comjnl/28.1.82 · doi:10.1093/comjnl/28.1.82
[9] Scott J., Combinatorial Programming, Spatial Analysis, andPlanning (1971)
[10] Simeone B., Atti Giornate di Lavoro AIRO pp 57– (1978)
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.