Spatial Cluster Detection Imposing Constraint on Shape Complexity in Networks

Abstract

: Point event cluster detection in networks have been proposed recently; they are suitable for analyses based on detailed location information, as they can describe the micro-space variation of locations of point events at the street level. However, the previous methods lack the flexibility to control the shapes of detected clusters; one can only detect ‘ circle-like ’ compact clusters that might include links where point events are scarcely distributed, and the other can only detect complex-shaped clusters that are difficult to interpret their causes. This paper proposes a shape complexity index in networks and a new cluster detection method imposing constraint on the shape complexity based on the proposed index. The application revealed that the proposed method succeeds in controlling the shape complexity of detected clusters in networks.