Efficient Multilevel Spatial Co-location Pattern Discovery Based on Density-Wise Clustering and MC-Hash Structure

Spatial co-location pattern (SCP) mining is a significant research direction in data analysis and has been proven in various fields. However, currently SCP mining still faces many challenges, e.g., dealing with heterogeneous distribution data, the expensive computational cost required for mining multilevel SCPs. To address the two mentioned challenges, a new method applied density-wise clustering is proposed to discover multilevel SCPs in heterogeneous distribution data. The data first is split into clusters based on both density levels and distribution. Then, Bron-Kerbosch with a pivot is adopted to list all maximal cliques (MCs) from each cluster. Then, the MCs are set in to a MC-hash structure to enhance the SCP mining process. The proposed method is examined and compared with other two efficient methods on both synthetic and real-world data. The experimental results show that our method not only effectively addresses the mentioned challenges but also generates noteworthy results.