RCP Mining: Towards the Summarization of Spatial Co-location Patterns

Abstract

AbstractCo-location pattern mining is an important task in spatial data mining. However, the traditional framework of co-location pattern mining produces an exponential number of patterns because of the downward closure property, which makes it hard for users to understand, or apply. To address this issue, in this paper, we study the problem of mining representative co-location patterns (RCP). We first define a covering relationship between two co-location patterns by finding a new measure to appropriately quantify the distance between patterns in terms of their prevalence, based on which the problem of RCP mining is formally formulated. To solve the problem of RCP mining, we first propose an algorithm called RCPFast, adopting the post-mining framework that is commonly used by existing distance-based pattern summarization techniques. To address the peculiar challenge in spatial data mining, we further propose another algorithm, RCPMS, which employs the mine-and-summarize framework that pushes pattern summarization into the co-location mining process. Optimization strategies are also designed to further improve the performance of RCPMS. Our experimental results on both synthetic and real-world data sets demonstrate that RCP mining effectively summarizes spatial co-location patterns, and RCPMS is more efficient than RCPFast, especially on dense data sets.KeywordsApproximation StrategyFrequent ItemsetCompression RateRepresentative PatternCover RelationshipThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.