Spatio-temporal cluster detection of point events by hierarchical search of adjacent area unit combinations

Several methods for detecting spatial or spatio-temporal clusters of point events, based on the "space scan statistics" presented by Kulldorff and Nagarwalla (1995), have been proposed. In most of them, it is implicitly assumed that the propagation process starts from a certain point, and an at-tempt is made to extract a circular-like spatial region as a cluster. Howev-er, when analyzing socio-economic phenomenon in small areas, the as-sumption of an isotropic propagation process is not always appropriate, since the shape of a transport network and the non-homogeneity of attrib-utes in space affects the speed of propagation significantly. In this paper, a method for detecting flexible-shaped clusters by searching combinations of adjacent area units hierarchically is proposed. The applicability of the pro-posed method to the detection of spatio-temporal clusters is tested on transaction location and time data of real estate in Tokyo.