Abstract
With the growing interest in commercial trend analysis, mining inter-transaction association rules in transaction databases has become an important data mining research. Mining inter-transaction association rules create new challenging problems since they involve date items in multiple transactions.Previous studies adopt an Apriori-like generation-and-test approach utilized to mine intra-transaction associations. In this study, we propose graph-based algorithms to discover specific large-k inter-transaction itemsets and maximal large inter-transaction itemsets (a maximal large inter-transaction itemset is not a subset of any large inter-transaction itemset.). Most users are interested in some specified rules, hence, discovering specific large-k inter-transaction itemsets efficiently will be helpful to satisfy users' requirements. Moreover, finding maximal large inter-transaction itemsets is another solution for mining all large inter-transaction itemsets. Our performance shows that our algorithms perform well, especially for those potentially long large inter-transaction itemsets. In addition, we analyze the approaches on mining various patterns from transaction or sequence databases. As a result, a classification of this problem into five classes is formed.