An Efficient Algorithm for Mining Closed Frequent Inter-transaction Itemsets

Many algorithms have been proposed recently for finding inter-transaction association rules, which represent the relationships among itemsets across different transactions. Since numerous frequent inter-transaction itemsets will be generated, mining closed frequent inter-transaction itemsets can speed up the mining process.
Therefore, in this thesis, we propose an algorithm, ICMiner (Inter-transaction Closed patterns Miner), to mine closed frequent inter-transaction itemsets. Our proposed algorithm consists of two phases. First, we convert the original transaction database into a set of domain attributes, datset, for each frequent item. Second, we enumerate closed frequent inter-transaction itemsets by using an itemset-datset tree, ID-tree. Mining closed frequent inter-transaction itemsets with an ID-tree, we can avoid costly candidate generation and repeatedly support counting. The experimental results show that our proposed algorithm outperforms the FITI and ClosedPROWL algorithms by one order of magnitude.