Hierarchical clustering through spatial interaction data. The case of commuting flows in south-eastern France

Abstract

Regional scientists' methods to partition space in functional areas meet complex system analysts' methods to detect communities in networks. A common concern is the detection of hierarchical sets of clusters representing underlying structures. In this paper modularity optimization in complex networks is compared to polarized functional area definition through dominant flows. Different approaches to the significance of dominant flows are also tested, namely threshold and Multiple Linkage Analysis approaches. Both methods are applied recursively in order to obtain a hierarchical clustering of municipalities in the PACA region (France) based on commuting flows in 1999. The comparison focuses on the geographical meaning of the results of the analyses. Modularity optimization and dominant flow results agree in many points and highlight the inadequacy of official methods integrating administrative boundaries in functional area definition. When they differ, they offer complementary views on the urban structure of the PACA region. © 2011 Springer-Verlag.