基于带语义差别的模糊Taxonomy的交易数据库关联规则聚类  

Association Rules Clustering Method Based on Fuzzy Taxonomy with Semantic Information

作　　者：沈斌[1,2] 姚敏[2] 刘艳彬[1] 

机构地区：[1]浙江大学宁波理工学院,宁波315100 [2]浙江大学计算机科学与技术学院,杭州310027

出　　处：《情报学报》2010年第2期246-253,共8页Journal of the China Society for Scientific and Technical Information

基　　金：国家自然科学基金资助项目（60533040,60525202）;浙江省自然科学基金重点项目（N0.Z104267）;浙江大学宁波理工学院科研启动基金资助项目.

摘　　要：关联规则聚类是大量关联规则的一种有效组织方式，本文针对基于商品分类信息的规则聚类方法存在的不足进行了改进，同时考虑了不同层次间的项目语义差别，以及具有不同隶属度的项目细致语义差别，将商品分类树改进为模糊Taxonomy的有向无环图结构，该结构可以处理一个项目同时属于多个父结点的情况。我们充分考虑了有向无环图的性质，提出了带细致语义差别的模糊Taxonomy结构构建方法和相应的规则距离计算方法，其中，规则距离计算过程中的项集距离计算方法无需计算最佳匹配，因此，具有较小的时间开销。规则距离计算和聚类可视化试验结果表明了该方法的可扩展性和有效性，在规则的聚类计算上取得了较为满意的结果。Association rules clustering is an efficient organizing method for massive association rules. According to the shortcomings of existing association rules clustering method which is based on taxonomy information, this paper proposes an improved method based on fuzzy taxonomy with semantic information. This new method not only considers the semantic difference between different levels, but also the nuance between different items which have different degrees belonging to their common ancestor node. At the same time, fuzzy taxonomy with semantic information also can well deal with the situation that one item belongs to multiple ancestor nodes, which cannot be managed in the existing taxonomy information based method. An establishing method of fuzzy taxonomy with semantic information and its corresponding rule distance computation method are proposed. Among these procedures, the computation method of item distance does not need to compute the best matching, so it has a low time cost. The experiments of computation of rule distance and visualization of clustering show that our method is efficient and can achieve satisfied results.

关 键 词：关联规则 聚类 模糊Taxonomy 语义信息 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]