一种新的空间多维关联规则模型与算法

New Spatial Multi-Dimensional Association Rule Model and Its Algorithm

空间对象具有自相关、连续性、多尺度等特点,导致空间关联规则挖掘与传统的统计关联规则挖掘不同,不存在统计的“事务”,挖掘更加复杂。本文用基于空间相关的影响域来创建“空间事务”,以代替传统关联规则挖掘中的事务,建立了一种新的应用于挖掘空间多维数据的空间多维关联规则模型(Spatialmultidimensionalassociationrulesmodel,SMARM)。设计并实现了一种新的挖掘算法SMARBIA,用基于影响域、空间支持度等剪枝技巧,克服了空间多维关联规则挖掘过程中候选项目集庞大的困难。实验表明,该算法能有效地减少候选项目集而获得较好的性能。

Spatial association rule mining based on spatial objects is more difficult than relational association rule mining because the spatial object attributes are spatially auto-correlated, continuous and multidimensional. It is more difficult to define transactions in traditional association rule mining. This paper establishes a new spatial multi-dimensional association rules model (SMARM) for multi-dimensional spatial data mining, where spatial transactions are defined by a notion of impact zone based on spatial autocorrelation and replaced by a traditional transaction definition. A new mining algorithm (SMARBIA) is realized. The algorithm can avoid enormous candidate items in mining process by pruning techniques based on impact zone and spatial support. Finally the experiment shows that it can decrease the number of candidate items.