时态数据库中非数值型属性周期规律的研究

MINING PERIODICITY RULES OF ATTRIBUTE OF NON-NUMBER TYPE AND ASSOCIATION RULES FROM TEMPORAL DATABASE

时间是数据本身固有的属性,将时间约束加在关联规则中能更好地说明事实.本文介绍的方法能够提取时态数据库中带时态信息的关联规则,而且能够计算时态数据库中某个非数值型属性(项)的周期,并通过执行改造了的Apriori算法提取该属性的周期规律.本文通过选取两个时间粒度,对时态数据库中的时间区间进行了两次划分和标记.第一次划分和标记的目的是计算选择出的某非数值型属性的周期;第二次划分和标记的目的是离散化时间区间,用标记集合代表原时间区间,进而根据标记集合求交的结果得到带时态信息的频繁项集.采用标记集合求交的方法能够使得Apriori算法的迭代迅速收敛,提高算法执行效率.

Because time is the inherent factor of data, the association roles will be better in illuminating the fact when adding temporal constrain. This paper proposes an algorithm for mining the association roles with temporal constrain from temporal database. In addition, our algorithm can compute the periodicity of an attribute（item）of non- number type of temporal database. By using the adaptive Apriori algorithm, our algorithm can mine the periodicity rules of this item of nonnumber type. Our algorithm divides and marks the valid time item of temporal database two times by choosing two time granularity respectively. The first operation of dividing and marking trys to calculate the periodicity of the item of non - number type that we have chosen. And the purpose of the second operation is discretizing the valid time item and regarding the sign set as the representation of the valid time item, then our algorithm gets the frequency itemset with temporal information according to the results of intersecting the sign set. This feature of our algorithm can make the Apriori algorithm more efficient by significantly improving the speed of constringency of iterations of Apriori.