Ming Non-redundant Associations From the Frequent Concept Sets on FP-tree

The classical algorithm for mining association rules is low efficiency. Generally there is high redundancy between gained rules. To solve these problems, a new algorithm of finding non-redundant association rules based on frequent concept sets was proposed. The Hasse graph of these concepts was generated on the basis of the FP-tree. Because of the restriction of the support most Hasse graphs have lose lattice structure. During building process of the Hasse graph, all nodes were formatted according to the index of items which were found in the frequent-item head table. At the same time these nodes were selected by comparing supports. In the Hasse graph, the intention of node is frequent itemset and the extension of node is support count of this item set. And the non-redundant association rules were gained by scanning the leaf nodes of the graph. The simulation shows the feasibility of the algorithm proposed.