Researches And Applications On Association Rules Mining With Multiple Minimum Supports

Data mining refers to extracting or "mining" knowledge from large amounts of data, which is deemed to one of a foreland of information system and promising cross-subject. Association rules mining is one of the most important part in data mining. As a new research field of association rules mining, mining association rules with multiple minimum supports is regarded as an important research for mining association rules in rare items.We analyzed the properties of multiple minimum supports association mining based on analyzed other algorithms of association rules mining with multiple minimum supports, and propose a new tree structure Set-enumeration Binary Tree(SEB-tree),to compress and store the mining dataset. An efficient algorithms-MSTApriori, was designed for mining minimum support association rules based on SEB-tree. The MSTApriori algorithm is more efficient than the typical MSApriori algorithm, when the dataset is big or has more duplicate transactions. We also propose a term Reference Attributes Set of items for setting items' minimum support, and successfully use BPNN to assist user setting items' minimum support, which is the other innovation.Our performance study shows that the algorithm is efficient on typical mushroom dataset and synthetic data. In addition, we used the algorithm on dataset of score and electronic police, and abstracted some rules for assisting each department to make advantaged decision.