关键词: 流数据, 聚类, 数据挖掘

Abstract: As data stream is dynamic, its model will change with time. If the change in data stream’ model can be detected and reported detailedly in time, it will help the user make out better strategies. Aiming at this requirement, this paper proposed a new strategy based on CFT, by which not only the change can be found, but also the difference between the new cluster model and the old cluster model can be discovered in detail. The time complexity of the strategy is O(K2), and the experiments show that the change can be detected accurately and reported in a intuitive way.

Key words: data stream, cluster, data mining