Abstract: This paper analyzes some features of discovering frequent patterns on large set of transactions. On this basis, it gives an improvedalgorithm (called AprioriAdjust) with the development of two techniques: (1) applying the compressive processing of the transaction set for miningsequential patterns; (2) presenting a method based on the statistical mechanism to evaluate the sequence’s support, in which the convergence ofaverage value of the support in the whole procedure is considered, so as to effectively prune the candidate set of the frequent patterns. Furthermore, itdiscusses the results of the experiments. Compared to the AprioriTID algorithm, it shows that the AprioriAdjust algorithm is more efficient andscalable over the large sets of transactions.

摘要： 分析了在大型事务集合发现频繁模式的特点,给出了一个改进算法(称为AprioriAdjust算法),同时,发展了两项技术:(1)提出了事务集合的压缩过程。(2)展示了一种基于统计模式的方法来评价序列的支持度,在该方法中,考虑了整个过程中每一趟的支持度均值的收敛性,以进行有效的候选频繁集的剪枝。此外,还讨论了实验结果。比较AprioriTID算法,研究展示,AprioriAdjust算法在进行大型事务集合的挖掘时更有效,可扩展性更强。

关键词: 数据挖掘, 序列模式挖掘, 最大频繁项目集, 统计模型

CLC Number: 

• TP311.13