Incremental Mining of Spatial Co-Location Patterns Based on the Fuzzy Neighborhood Relationship

摘要

A spatial co-location pattern is a set of spatial features frequently co-located in nearby geographic spaces. Due to the spatial database is constantly changing as time goes on, the incremental mining of prevalent co-location pattern algorithms have been proposed in the literature. And focusing on the ignorance of the proximity level between instances, the co-location pattern mining based on fuzzy neighborhood relationship (FNR) has also been studied. However, the problem of incremental mining of prevalent co-location patterns based on fuzzy neighborhood relationship on the dynamic databases has not been addressed. In this paper, based on FNR, by capturing the changed (added and decreased) fuzzy neighborhood relationships, we define the incremental fuzzy participation index for measuring the prevalence of the changed co-location in the updated data sets, and design the algorithm of incremental mining of prevalent co-location patterns based on FNR (the IMPCP-FNR algorithm). Extensive experiments are conducted and demonstrate that, by compared to the naive method that re-discovers the prevalent co-locations on the whole updated data sets, our purposed algorithm is more efficient.