The FP-Tree Algorithm Used for Data Stream

FP-growth is a classical algorithm in frequent pattern mining, which is often used in static data mining. Some researches have been done on using FP-growth algorithm to analyze streaming data. However, the double-scan-of-database manner in FP-tree creation is a serious bottleneck in streaming data analysis. Sliding window technique could solve this problem in certain degree, but it still can lead to the inaccuracy of FP-tree creation, which may impact the consequent data mining. In this paper, a new FP-tree algorithm is presented for streaming data, which creates the FP-tree by a single-pass scanning (SPSFP) throughout the database. Compared with the traditional FP-tree creation, the new method scans the database only once and doesn´t need to store the whole data set into memory, which not only saves the memory space but also makes it possible to mine frequent pattern accurately in streaming data. Furthermore, the time cost of the new algorithm is almost equivalent to the traditional one.