Sequential Patterns Mining Scaling with Data Stream Based on LSP-tree

We present a new method of mining sequential patterns in data stream based on a fast bitmap method. In recent years data stream emerges as a new data type in many applications. When processing data stream, the memory is fixed, new stream elements flow continuously. The stream data can not be paused or completely stored. We developed a LSP-tree data structure to store the discovered sequential patterns. To be suitable for the time-changing stream data, a time-tilted window is applied to scale with continuously increased LSP-tree. Experiments on a set of large data stream demonstrate the utility of this algorithm.