Abstract: Frequent data updates in data stream make the whole database re-mineing impossible. This paper proposes maintaining generator representation and border sets based on recent data in stream. Representation changes aroused by data updates can be detected in those border sets. Then updating work is confined to those item sets related with updated transactions and achieves good results.

Key words: Generator, Data stream, Data mining

摘要： 数据流中频繁的数据更新使得重新挖掘整个数据集显得比较困难。该文提出了在数据流中，基于最近数据的动态维护Generator表示方法。通过界定边界项集，使得由数据更新可能引起的项集变化能在边界集中被检测到，而无须保存所有频繁集，使处理限定在仅与更新相关的项集范围之内，取得了较好效果。

关键词: Generator, 数据流, 数据挖掘

CLC Number: 

• TP311