减量式频集快速维护算法研究与实现

ON RAPID DECREMENTAL MAINTENANCE ALGORITHM FOR FREQUENT ITEMSET AND ITS IMPLEMENTATION

在频集更新算法的研究中,关于数据集减量式的更新算法研究比较少。提出一种最小支持度和置信度不变,从事务数据库中删除一个事务数据集后,如何高效地生成变化后的事务数据库频集的算法。算法在如何充分利用以往挖掘过程中的信息,避免多次扫描数据集以及如何减少候选集的规模等方面进行了研究,给出了算法的实现。通过对实验结果的性能对比分析,表明算法是可行、有效的。

In study of the updating algorithm for frequent itemset,little research was made on updating algorithm of datasets in decremental form. This paper provides an algorithm for efficiently generating frequent itemset of the changed transaction database with the minimum support degree and confidence degree keeping unchanged while a transaction dataset is deleted from the transaction database. The algorithm explores various aspects encompassing how to make the best use of known information from previous mining processes, how to prevent from repeatedly scanning datasets and how to decrease the size of candidate sets, etc. The implementation of the algorithm is also given. It is manifested through the comparative analysis of the performances of experimental results, the algorithm is feasible and effective.