Dataset filtering based association rule updating in small-sized temporal databases

Association rule mining can uncover the most frequent patterns from large datasets. This algorithm such as Apriori, however, is time-consuming task. In this paper we examine the issue of maintaining association rules from newly streaming dataset in temporal databases. More importantly, we have focused on the temporal databases of which storage are restricted to relatively small sized. In order to deal with this problem, temporal constraints estimated by linear regression is applied to dataset filtering, which is a repeated task deleting records conflicted with these constraints. For conducting experiments, we simulated datasets made by synthetic data generator.