A parallel algorithm for regional co-location mining based on fuzzy density peak clustering
Xiwen Jiang, 丽珍 王, Vanha Tran
Abstract
Regional co-location pattern mining (RCPM) is designed to discover co-location patterns that exist within some local regions to address the patterns that cannot be found globally. Traditional RCPM techniques use geometry with well-defined boundaries as the regions of prevalent co-locations. However, the regions should not have such determined bounds. Moreover, data distribution is another important feature of a region, and this feature should also affect region selection. Based on the above considerations, we introduced density peak-based clustering (DPC) and proposed a novel density metric, combining with the fuzzy set theory and k-nearest neighbor distance to design an applicable paralleled RCPM algorithm. Experiment results show that our method can mine more meaningful results, and parallelization improves our algorithm efficiency. On real data, the speedup ratio under two threads reached 1.89.