关键词: 序列模式, 数据流挖掘, 滑动窗口

Abstract: Sequential pattern mining is one of the most important tasks of data mining and has broad applications. Sequential pattern mining has been studied extensively in static databases. However, the study of sequential pattern mining based on data streams is not very deep. Stream data has the characteristic of unlimited flow, it can not save all the data, and people usually are interested in the sequential patterns in recent time period, accordingly it introduces one effective method combining with sliding window technique for mining sequential patterns from data streams: FPM-SW algorithm（Frequent Pattern Mining-Sliding Window）. It uses three data structures（PatternTable, CountTable and Ta-tree） to handle the complexities of mining frequent sequential patterns in data streams. FPM-SW algorithm uses CountTable structure to preserve the past potential frequent sequences, considering that in some cases the countTable uses too much memory, the algorithm also combines a CountTable compression techniques to reduce memory footprint. The excellence of the algorithm is that it can maximize the reduction of the number of false positive. Experimental results show that FPM-SW has higher accuracy.

Key words: sequential patterns, data stream mining, sliding window