OSCRM:一个基于本体的空间Co-Location规则挖掘框架

OSCRM:A Framework of Ontology-Based Spatial Co-Location Rule Mining

空间co-location挖掘是空间数据挖掘的一个重要方向.但现有的挖掘算法很少甚至不考虑挖掘领域的背景或约束知识,挖掘到的大量co-location规则给决策者带来了极大的困扰.因此,提出一种基于本体的交互式空间co-location规则挖掘框架OSCRM(ontology-based spatial co-location rule mining).首先,OSCRM提供了基于本体的用户领域知识表达机制;然后,OSCRM提供了本体之上的强大的公式系统,使用户可以方便地表达指导挖掘方向的领域背景或约束知识;接着,OSCRM提供了2个经典的空间co-location挖掘算法,算法实现中充分利用了用户提供的公式集进行过滤处理;最后,OSCRM还提供了一种交互式的后处理机制(二次挖掘机制),进一步地减少最终规则的数量.使用实际数据的实验表明OSCRM不仅是一个方便、实用的领域驱动空间co-location挖掘框架,更为重要的是规则过滤效率达到了99.9%.

Spatial co-location mining plays an important role in spatial data mining.But existing mining algorithms seldom take into account the background of mining domain or constraint knowledge,the huge amount of mined co-location rules brings the decision-makers great frustration.In this paper,we propose a framework called OSCRM which is an interactive spatial co-location mining framework based on the ontology.Firstly,OSCRM provides an ontology-based expression on usesdomain knowledge;secondly,OSCRM contains a powerful formula system which can easily represent domains background or constraint knowledge;and then OSCRM uses two classic colocation mining algorithms in which we make full use of formula sets provided by users for rule pruning to mine coarse rule sets;finally,OSCRM has an interactive post-processing step(the secondary mining)to reduce the number of rules further more.The results of our experiments show that OSCRM is not only a convenient and useful domain-driven spatial co-location mining framework,but also its efficiency of rule filtering can reach 99.9%.