A Spatial Scan Statistic to Detect Spatial Communities of Vehicle Movements on Urban Road Networks

Abstract

Identifying spatial communities in massive vehicle trajectory data greatly facilitates the understanding of spatial interactions in a city. However, it is still challenging to identify irregularly shaped and statistically significant spatial communities in vehicle movements. To overcome this challenge, we develop a spatial scan statistic based on ant colony optimization. A spatially embedded network was constructed with road segments as nodes and the numbers of vehicle trips between the road segments as weights. To evaluate the statistical significance of spatial communities, we first defined a random graph with a given expected strength sequence, and then, constructed a new likelihood ratio test statistic. To detect irregularly shaped spatial communities without a brute‐force search, a significance test was first used to identify road segments highly interacting with other road segments, and then, a contiguity‐constrained ant colony optimization was employed to combine these road segments to form spatial communities. The statistical significance of the spatial communities was evaluated using Monte Carlo simulation. Experiments on both the simulated and observed trajectories showed that the proposed method outperforms the five state‐of‐the‐art methods in detecting irregularly shaped spatial communities and ensures that the detected spatial communities are statistically significant.