Abstract: Mining frequent itemsets is a basic task of the data stream mining. Recently many approximate algorithms can mine frequent itemsets over data stream. However， these algorithms still can not efficiently reduce space and time cost. To improve the efficiency of mining frequent itemsets over data stream， matrix is imported as the synopsis data structure and a new algorithm of mining frequent itemsets is presented. Finally， experiments prove the efficiency of this algorithm.

Key words: data stream, data mining, frequent itemsets, matrix

摘要： 挖掘频繁项集是挖掘数据流的基本任务。许多近似算法能够有效地对数据流进行频繁项挖掘，但不能有效地控制内存资源消耗和挖掘运行时间。为了提高数据流频繁项集挖掘的时空效率，通过引入矩阵作为概要数据结构，提出了一种新的数据流频繁项集挖掘算法。最后通过实验证明了该算法的有效性。

关键词: 数据流, 数据挖掘, 频繁模式, 矩阵