Abstract: Along with the sharp increment of the information, mining frectuent itemsets gradually becomes a hot point in recent years.日'-Stream is a classic algorithm for mining frequent itemsets at multiple time granularitics.But the weakness is the contradiction of the massive data and the limited memory in a certain time which will lead to the result that the algorithm can not be used in high speed data stream mining. This article proposed a improved FP-Stream algorithm based on vertical and Dif-bits compression data format

Key words: Data stream, Frectuent itemsets, FP-Stream, Vertical format, Dif-bits data-compression