A Clique-Querying Mining Framework for Discovering High Utility Co-Location Patterns without Generating Candidates

ACM TRANSACTIONS ON KNOWLEDGE DISCOVERY FROM DATA(2024)

引用 0|浏览29
暂无评分
摘要
Groups of spatial features whose instances frequently appear together in nearby areas are regarded as prevalent co-location patterns (PCPs). Traditional PCP mining ignores the significance of instances and features. However, in reality, these instances and features have different significance, the traditional PCPs may not sufficiently expose knowledge from spatial data. This study focuses on discovering high utility co-location patterns (HUCPs) in which each instance is assigned a utility to reflect its significance. To filter HUCPs, an adaptive utility participation index (UPI) is designed. Unfortunately, the UPI does not hold the downward closure property. The performance of mining HUCPs is very inefficient since unnecessary candidates cannot be early pruned. Thus, an efficient clique-querying mining framework is devised without generating candidates. This framework first divides neighboring instances into cliques, then compacts these cliques into a hash table structure. Next, the adaptive UPI of any patterns can be quickly calculated based on their participating instances that are obtained by executing a querying scheme on the hash table. Finally, HUCPs are filtered efficiently. The effectiveness and efficiency of the proposed method are proved in both theory and experiments to make a promise that the patterns mined are more meaningful and the mining performance is significantly improved compared to the previous methods.
更多
查看译文
关键词
High utility co-location pattern,spatial clique,participating instance,querying scheme
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要