Research On The Algorithm Of Mining Frequent Itemsets From Uncertain Data Based On The Tree

With the development of society, more and more data are collected by people. Sometimes they are partly lost or with the noise. But the old algorithms couldn't mine the frequent itemset from the uncertain data effectively, we need the new one to handle these problems.In this paper, we discuss the UF-growth algorithm firstly, and test its performance (compare it with the U-Apriori algorithm). But there are some problems: consumption memory excessively and take too much time. So we propose two methods to solve the difficulties,and three new algorithms:(1)To reduce the memory consumption and increase the chance of path sharing,we discretize and round the expected support,and propose the"LUF-growth".The experiments shows it is effective,and better than UF-growth.(2)When building the UF-tree,we improve the header table to save the time,and propose the"UFT-growth". The experiments shows it is effective and improve the efficiency.(3)When improving the UF-growth by these two methods, we propose the"LUFT-growth". The experiments shows the LUFT-growth save more time than others.