时态数据库中增量关联规则的挖掘

MINING THE INCREMENT ASSOCIATION RULES IN TEMPORAL DATABASE

时态数据库 (tem poral database)中的时态数据是数据信息中重要的一类 ,此类数据中存在着大量未知的涉及数据的变化趋势及数据增量之间相互关系方面的知识 .给出了时态数据库中的相邻关系 (adjacency)和增量(increm ent)的定义 ,提出了基于定量属性离散化及编码化的增量关联规则的基本模型 (model of discretization &code- based increment association rules) ;并针对时态数据库中的属性大多为连续定量属性的特性 ,比较了“部分完备法”(partial com pleteness)、基于分类信息熵的离散化 (entropy- based discretization)、C4.5、基于粗集理论的全局离散化 (rough sets- based full discretization)及其改进等多种定量属性的离散化方法 ,结合定性属性的挖掘算法 ,提出了具有时间约束的增量关联规则挖掘算法 TIDM;最后 ,讨论了互斥属性集的概念、基本特性及处理互斥属性集的基本方法 。

Temporal Database is one of the important data information systems. There is some useful unknown knowledge concering the tendency and the relation of data increment. In this paper the definitions of adjacency, increment, collision set are given and the model of discretization & code based increment association rules are proposed. Some discretization methods for quantitative attributes are compared, which includes partial completeness, entropy based discretization, C4.5, and rough sets based full discretization. Finally, the algorithm TIDM that joins an ameliorative method RBWAD of rough sets based full discretization with general data mining is discussed.