Meta-PCP: A concise representation of prevalent co-location patterns discovered from spatial data

Abstract

A prevalent co-location pattern (PCP), which is a group of spatial features whose spatial instances frequently appear together in nearby geographical areas, can expose valuable information and knowledge that can be applied to many fields. In traditional PCP mining, to filter interesting PCPs, a minimum prevalence threshold is employed. This threshold should be set to a small value to obtain as much information and knowledge as possible from spatial data sets. However, at this time, not only too many redundant patterns are found, but also mining efficiency is extremely low and memory space consumption is very high. To solve this self-contradiction, this paper proposes a new concept that is called meta-prevalent co-location pattern (meta-PCP). Meta-PCPs can eliminate redundant information and concisely represent the mining result. Although meta-PCPs are a lossy representation of all PCPs, they can be controlled by users according to their application scenarios. Moreover, a query-based mining algorithm is designed to improve mining performance when the prevalence threshold is set to very low. This algorithm discovers meta-PCPs without generating candidates (to improve efficiency) and does not collect and remain co-location instances of each pattern (to reduce memory consumption). The comprehensive experimental results on both synthetic and real data sets show that the proposed method is effective and efficient.