Abstract: As one of the most important problems in data stream mining，the frequent patterns mining with a sliding window is widely researched and used in many fields.Exiting algorithms need process all elements in the data stream，whereas users only focus on several aspects of things.So inspired by the MFI-TransSW algorithm，a new algorithm based on transaction-sensitive sliding window is proposed in this paper，in which a sequence of bits is used to implement the sliding window operation.In addition，a mechanism of filtering frequent items，which decreases the memory usage and improve the efficiency of processing，because of the reduction of items retained in memory.Furthermore as space complexity is independent to the size of sliding window and the value range of elements，this method is specially applicable to discovery of data with a wide range of values in a long period.The analysis and experiments show the feasibility and effectiveness of the algorithm.