Mining Minimal Non-Redundant Association Rules Using FP-Trees

Two problems exit in the association rules mining, firstly the mining efficiency is low, secondly, the quantity of rules is large and the amount of interesting rules is small. In this study a solution is proposed to mine the minimal non-redundant association rules using frequent-pattern tree (FP-tree). The work mainly includes the following aspects.The AClose algorithm which is the classical algorithm of mining minimal non-redundant association rules is analyzed. The mining algorithm based on FP-tree is proposed to improve the low efficiency of the AClose algorithm.Frequent smallest itemsets and referenced properties are proposed for minimal non-redundant association rules mining based on FP-tree. Furthermore, algorithm called MFF is presented to mine frequent closed itemsets and frequent smallest itemsets.A result tree based on FP-tree is constructed to store all frequent closed itemsets and frequent minimal itemsets aiming to improve the efficiency of mining non-redundant association rules. The method of mining minimal non-redundant association rules using result tree is presented.In the processes of implementing algorithm, four strategies are applied to improve the efficiency of traversing nodes. They are, redefining the structure of the FP-tree, converting multiway tree to binary tree, applying a data item of frequent smallest itemsets for the node structure of result tree, building hash index for new FP-tree structure and result tree.The theory analysis and experiments validates the feasibility of the proposed solution. Experiments show that MFF outperforms over AClose.Finally, further work is depicted for mining non-redundant association rules using FP-tree.