Abstract: High dimensional large data has posed great challenges to most existing algorithms for frequent patterns mining. This paper decomposes the mining task into two parts: mining short frequent itemsets and long frequent itemsets, and proposes a new algorithm, i.e., inter-transaction, to find all long frequent closet patterns in large high dimensional dataset. The new algorithm utilizes the characteristic that the intersection of long transactions is usually a very short itemset, and can find long closet patterns directly via intersecting relevant transactions. In addition, the algorithm adopts a new pruning strategy to cut down search space and optimizes the performance of intersection of transactions. Experiments on synthetic data show that this method achieves high performance in large high dimensional dataset.

Key words: large high dimensional dataset, frequent closet pattern, pruning strategy

摘要： 高维大数据集对现有的数据挖掘算法提出了挑战。该文把挖掘任务分解为挖掘频繁长模式与短模式2个子问题，提出一种在高维大数据集中挖掘长项集的算法，即inter-transaction。该算法利用了高维数据中长事务相交迅速变短的特性，通过事务的交集运算直接得到长闭合模式，同时采用新的减枝策略，优化了事务交集运算的方法。实验表明，该方法对高维大数据集非常有效。

关键词: 高维大数据集, 频繁闭合模式, 减枝策略

CLC Number: 

• TP311