Mining Accurate Top-K Frequent Closed Itemset from Data Stream

Frequent Closed Item set mining on data streams is of great significance. Though a minimum support threshold is assumed to be available in classical mining, it is hard to determine it in data streams. Hence, it is more reasonable to ask users to set a bound on the result size. Therefore, a real-time single-pass algorithm, called Top-k frequent closed item sets and a new way of updating the minimum support were proposed for mining top-K closed item sets from data streams efficiently. A novel algorithm, called Can(T), is developed for mining the essential candidate of closed item sets generated so far. Experimental results show that the proposed the algorithm in this paper is an efficient method for mining top-K frequent item sets from data streams.