A variable size sliding window based frequent itemsets mining algorithm in data stream
Haiqing Li, Lang Wang · AIP conference proceedings · 2017
Due to the unpredictability and the concept drift character of the data stream, the traditional sliding window is difficult to adapt to frequent itemsets mining in data stream. A new variable sliding window based VSW-SCPS algorithm is proposed. The algorithm maintains a tree structure of SCPS-tree in memory, which is the storage structure of sliding window. When data flowing in, the SCPS-tree will adjust dynamically, the window size will be adjusted by the detection of concept drift according to the result of FP-growth mining algorithm. Experimental results show that the proposed algorithm has good time efficiency.