关键词: 数据流挖掘, 频繁模式, 倾斜时间窗口

Abstract: A TWCT tree structure is proposed with tilted-time window framework embedded，which can maintain the complete set of frequent patterns at multiple time granularities.And this paper designs the sequential update and delete algorithms for the structure，which makes it can be saved in auxiliary storage in order to reduce the algorithms’ requirements of the main memory effectively.Taking advantage of this characteristic，TWCT-Stream，a frequent pattern mining algorithm in data stream，is proposed.And its pattern growth algorithm，TWCT-Growth，generates frequent patterns in lexical order suitable for the sequential updating of TWCT structure.Experiments have proved its memory requirements lower than the same kind of algorithms like FP-Stream and so on.

Key words: data stream mining, frequent pattern, tilted-time window