Detecting Spatial Communities in Vehicle Movements by Combining Multi-Level Merging and Consensus Clustering

Identifying spatial communities in vehicle movements is vital for sensing human mobility patterns and urban structures. Spatial community detection has been proven to be an NP-Hard problem. Heuristic algorithms were widely used detecting communities. However, the identified by existing heuristic are usually locally optimal unstable. To alleviate these limitations, this study developed a hybrid algorithm combining multi-level merging consensus clustering. We first constructed weighted spatially embedded network with road segments as vertices numbers of trips between weights. Then, jump out local optimum trap, new approach, i.e., iterative moving global perturbation, was proposed optimize objective function (i.e., modularity) until maximum modularity obtained. Finally, obtain representative reliable structure, clustering performed generate more stable structure set results. Experiments on Beijing taxi trajectory data show that method outperforms state-of-the-art method, constrained Leiden (Scleiden), because can escape from solutions improve stability structure. The reveal polycentric Beijing, which may provide useful references human-centric planning.