Mining Closed Frequent Itemsets in Sliding Window over Data Streams

As one of the most important problems in data streams mining, many studies have been done on mining closed frequent itemsets. However mining closed frequent itemsets in data streams has not been well addressed. In this paper, we design HCI-Mtree (Hash-based Closed Itemsets Monolayer tree) to maintain the complete set of current closed itemsets. In HCI-Mtree, the itemsets with the same frequency are linked to the same hash-based counter. To mining closed frequent itemsets in sliding window over data streams, we propose a novel approach HCFI (algorithm based on HCI-Mtree for mining Closed Frequent Itemsets). Vertical representation of transactions is utilized in our algorithm to save processing time and space consuming. Our experiments show that HCFI has good performance especially when the window size is large.