DETECTION OF INDUSTRIAL AGGLOMERATIONS THROUGH A PROBABILISTIC MODEL RELAXING CONSTRAINTS ON ADJACENCY

Abstract

Industrial agglomeration is of strong interest to not only academic researchers but also policymakers because Industrial agglomeration enhances corporate productivity. As a first step to understand phenomenon of industrial agglomeration, we need to reveal empirical characteristics of industrial agglomeration. Spatial cluster-detection analysis is one of the analyses understanding empirical characteristics of industrial agglomeration. A lot of cluster-detection methods have been proposed. However, there is little method that relaxes constraint on adjacency of geographical units that compose a spatial cluster. Constraint that requires exact adjacency may have significant impact on detected clusters and results of analyses using detected clusters. We propose a new cluster-detection method that relaxes constraint on adjacency of geographical units belonging clusters. Along the lines of model-based clustering, we assume spatial data arise through a probabilistic model. Employing Potts model on probabilistic model, we can relax constraints on adjacency. The results of case study show that our method can detect clusters consist of non-adjacent geographical units.