MFISW: A New Method for Mining Frequent Itemsets in Time and Transaction Sensitive Sliding Window

It is challenge to design an efficient summary data structure and an online approximation algorithms to limit the memory usage and the scan times in streaming data mining. In this paper, we present a CST(compressed suffix tree) structure to store arriving itemsets in the SC model. Then, our MFISW (mining frequent itemsets in sliding window) algorithm with the top-down traversal strategy can only scan data once to mine frequent itemsets in sliding window. Next, MFISW algorithm can update the mining result incrementally. Experiment shows that MFISW is efficient and scalable.