【摘要】 提出了一种针对不同时间序列间关联模式的发现方法,并阐述了以该方法为基础而构建的关联模式挖掘系统的结构.系统按步骤主要分成序列分割,模式聚类和关联模式挖掘三个部分.其中关联模式的发现基于共同作用机制的思想,即两个不同的时间序列之所以出现频繁的关联模式,必定存在某种共同机制的作用或者二者本身之间有某种因果关系.通过定义可靠度来度量作用强度,并以此作为阈值约束,大大降低了算法的复杂性,伸缩性好,产生的关联模式数量适当.将其应用于股市关联变动模式的发现验证了其有效性.更多还原

【Abstract】 This paper gives an algorithm for mining association patterns between two different time series, and describes the construction of a mining system. There are three main processes: segmenting, clustering and discovering association patterns in this system. The algorithm for association patterns discovery is based on the idea of common mechanism, i.e. an association pattern may be a result from a common mechanism or some causality between two series. The influence by the common mechanism more strong, the probability of two patterns more close to each other. Under the constraint of a reliability measuring the influence, the search space is enormously reduced, which brings on low complexity and fine extensibility. An instance shows its validity. \;更多还原