中文摘要：

提出一种矩阵加权关联模式支持度计算方法及其相关定理，给出矩阵加权项集剪枝策略，基于该剪枝策略提出一种基于项权值变化的矩阵加权关联规则挖掘算法MWAR-Miner（matrix-weighted association rules—miner）。该算法克服现有的项无加权和项权值固定条件下挖掘关联规则的缺陷，采用新的剪枝技术和模式支持度计算方法挖掘有效的矩阵加权关联规则，避免无效的和无趣的模式产生。以中文数据集CWT200g和英文数据集NTCIR-5为实验数据，理论分析和实验结果表明，与现有矩阵加权模式挖掘算法和基于无加权的挖掘算法比较，该算法挖掘的候选项集数量和挖掘时间明显减少，挖掘效率得到极大提高。

英文摘要：

This paper introduced a new computing method for support of matrix-weighted association patterns and the related theorems firstly, and then presented a new pruning strategy for matrix-weighted itemsets. Finally it proposed a novel algorithm of matrix-weighted association rules mining based on dynamic weight of item, MWAR-Miner（ matrix-weighted association rulesminer）. Overcoming the defects of the traditional mining methods, this algorithm adopted the new pruning means of itemsets and computing method of itemset support so as to discover valid matrix-weighted association rules, which could avoid the generation of ineffective and uninteresting patterns. Based on Chinese dataset CWT200g and English dataset NTCIR-5 for the experimental data, the theoretical analysis and experimental results show that the MWAR-Miner can evidently reduce mining time and the number of candidate itemsets compared with the existing mining algorithms based on matrix-weighted itemsets and unweighted item. In addition, its mining is more efficient than the available algorithms for comparison.