Research On Incremental Updating Of Maximum Frequent Itemsets And Maximum Length Frequent Itemsets

Now data mining is an important domain of database research. Mining association rules is an animated branch of data mining, too. At the same time, mining frequent itemsets is key technique and step of mining association rules. Existed algorithms on mining frequent itemsets are aimed at static data. But data changes in actual life at all times, such as online service and shopping detailed list. Moreover data in the internet is dynamic. Under the circumstance of the dynamic data, traditional algorithms on mining frequent itemsets have two defects. On one hand, frequent itemsets don't reflect truly the present data. On the other hand, because of large numbers of data, it is ineffective repeatedly to scan database. So two representative frequent itemsets are discussed after understanding variational data and analyzing current research on frequent itemsets.Firstly, in allusion to big support and small support, there are two modified incremental updating algorithms on mining maximum frequent itemsets. When support becomes bigger, the Bigger-SMFIU algorithm utilizes reversed judging methods to mine new maximum frequent itemsets. To the contrary, the Smaller-SMFIU algorithm solves the problem of smaller support. It scans previous maximum frequent itemsets from higher dimension to lower dimension and deals with them according to different situations. In the end new maximum frequent itemsets are presented.Secondly, an improved FP-tree is generated. In addition, a new algorithm based on the structure is advanced to mine maximum length frequent itemsets. At the same time databases are changing. It describes in detail incremental updated algorithms on maximum length frequent itemsets when databases become bigger and smaller.Finally, by means of contrasting experiments with existed algorithms, it certifies that these algorithms in this paper on mining maximum frequent itemsets and maximum length frequent itemsets are feasible and effective. Besides the execution efficiency of every algorithm is analyzed and contrasted.