Mining Frequent Patterns Based on IFP_Stream over Data Stream

Due to the characteristics of data stream, the memory space becomes insufficient at once when a large number of data flow into. Hence, how to reduce the memory consumption is one of the beneficial things. In this paper, we propose a new data structure IFP_Stream and an algorithm MFPBI based on this structure. The structure consists of a prefix tree with embedding a novel logarithmic tilted time window and a header table. Much memory space can be saved by merging the nodes automatically. The algorithm MFPBI could be performed only one-pass and the approximate results that exceed the user-defined support count would be produced. The experimental result shows that the memory using remains more steady.