Spatial cluster detection on detailed data without constraint of continuousness

Spatial cluster is the set of geographical units where concentration of events is observed. Spatial clusters provide useful information for under-standing mechanism and characteristic of socioeconomic activities. Sever-al methods have been proposed for cluster detection. However, there is no existing method relaxes a constraint on adjacency of geographical units that compose clusters. Constraint that requires exact adjacency may have significant impact on detected clusters, especially in the case of detailed data. In this study, we propose a new cluster detection method relaxes con-straints on shape and adjacency. Along the lines of model-based clustering, we assume spatial data arise through a probabilistic model. Employing Potts model on the probabilistic model, we can embed constraints on shape in the probabilistic model and relax constraints on geometric shape. The applicability of the proposed method is tested on case studies using mesh data of Japanese economic census.