关键词: 不确定性数据流, 最大频繁项集, 超集检测

Abstract: For large data bases, the number of frequent itemsets is huge and redundancy, and mining maximum frequent itemsets is more suitable because the scale of the output is much smaller. But traditional mining maximum frequent itemsets algorithm assumes the availability of precise data. Mining frequent itemsets from uncertain data streams is much more complicated than precise streams, and there is no research on mining maximum frequent itemsets over uncertain data streams until now. Therefore, aiming at the shortcoming, the paper proposes a maximum frequent itemsets mining algorithm TUFSMax. The algorithm adopts a decay window model to find frequent itemsets through calculating expected supports, and it uses a new method, called marking the tree nodes. By using the new method, algorithm TUFSMax can avoid super detection in the course of mining all of the maximum frequent itemsets, to save the detection time. Experimental results show that the proposed algorithm is efficient in time and space.

Key words: uncertain data stream, maximum frequent items, super check