Mining top-k fault tolerant frequent patterns with sliding windows in data streams

Mining frequent patterns over streaming data has become an important research focus field with broad applications. However, the real-world data may be usually polluted by uncontrolled factors. Fault-tolerant frequent pattern can express more generalized information than frequent pattern which is absolutely matched. Therefore, a novel single-pass algorithm is proposed for efficiently mining top-k fault-tolerant frequent pattern from data streams without minimum support threshold specified by user. A novel data structure is developed for maintaining the essential information of itemsets generated so far. Experimental results show that the developed algorithm is an efficient method for mining top-k fault-tolerant frequent pattern from data streams.