Approximation algorithm for frequent itemsets mining on uncertain dataset.

To improve the efficiency of frequent itemset mining upon uncertain dataset, addressing the issue of heavy computation cost of existing algorithms on judging whether to build sub header table for a certain item in the header table, this paper proposed an approximation algorithm called AAT-Mine, at the cost of losing a small portion of frequent itemsets, improved the overall mining performance. It evaluated the AAT-Mine algorithm using three datasets against classical and state of art algorithms. Experimental results show that AAT-Mine not only outperforms AT-Mine, MBP, IMBP, UF-Growth and CUFP-Mine in terms of running time, but also remains efficient memory usage. [ABSTRACT FROM AUTHOR]