数据库中动态关联规则的挖掘

关联规则能挖掘变量间的相互依赖关系,但是不能反映规则本身的变化规律．为此本文提出了动态关联规则．首先将整个待挖掘数据集按时间划分成若干子集,每个子集挖掘得到的每条规则分别生成一个支持度和一个置信度,这样每条规则在全集上就对应了一个支持度向量和一个置信度向量．通过分析支持度向量和置信度向量,不仅可以发现规则随时间变化的情况,也能够预测规则的发展趋势．本文还提出了两个挖掘动态关联规则的算法,且对他们做了比较．并给出了柱状图和时间序列两种方法分析这两个向量．最后给出了一个挖掘动态关联规则的应用实例。

Mining dynamic association rules in databases

Association rules may discover the relations between variables, but are unable to reflect the variation between relations. Consequently, dynamic association rule is introduced in this paper. In our method, the entire database is divided into a series of subsets in time field, and each rule from a subset has a measure of support and confidence. As a result, there are a vector of supports and a vector of confidences for each rule. It not only helps us discover the rule variation with time by analyzing the two vectors, but also predicts the future of a rule. Two algorithms for mining dynamic association rule are proposed in this paper, and a comparison of such two algorithms is also made. Subsequently, histograms and time series are described as ways for analyzing the two vectors. Finally, the effects of dynamic association rule are shown in an instance.