Research On He Algorithm About Mining Association Rule

Data Mining distills knowledge from a mass of data. So, it is also called Knowledge Discover from Database. It is a new research area involving several branchs of learning and containing many domains. Association rule is one of the most important domains among them, which finds the interesting relations between items or attributes of database. These relations are unknown and hide, i.e. it cannot be gotten with logic operations or statistic methods of traditional database operation techniques. So, mining association rule do not base on self-attributes but on co-appearance character among items of database.At the begin this paper firstly introduces some basic principal theories, directions of development and problems in the face of. And then, this paper presents the conceptions, classes and general thoughts of the algorithms about association rule. Among those, some association rule algorithms are discussed deeply.The interesting relations among items of dataset are released by association rule. Current research interesting in the association rule focuses on the algorithm about mining frequent closed itemsets. Based on the character of different dataset structure, the classic algorithm about mining frequent closed itemsets CLOSET+ need to adopt bottom-up physical tree-projection or top-down pseudo tree-projection strategy to get candidate frequent closed itemsets, and then checks it for obtaining frequent closed itemsets. So, the cost is high. This paper presents a novel mining frequent closed itemsets algorithm S-growth to mine FP-tree based on stack structure. This algorithm only needs to build one global FP-tree in the memory. And then, pushing and popping the stack do the process of mining FP-tree. The process of building conditional FP-tree or traveling FP-tree recursively is avoided. Especially, this algorithm can get complete and irredundant frequent closed itemsets directly.