A stream sequential pattern mining model

Stream is continuous, fast, dynamic and unlimited. Data stream cannot be stored in second storages for multiple scanning. In this paper, we propose a multiple level sequential pattern mining model, which is adapted to stream characteristic. It implements traditional mining algorithm over in-memory data to acquire accurate sequential patterns from data of ranges of stream. Besides, the model splits the memory into many levels to store sequential patterns under different minimum supports. In addition, this paper discusses the construction process of parameter optimization. Finally, a series of experiments is implemented to prove the effectiveness and efficiency of this model.