A dynamic layout of sliding window for frequent itemset mining over data streams

Mining frequent patterns over data streams is an interesting and challenging problem due to the emergence of new applications and limited resources of main memory and processing power. In this study, a novel sliding window based method for efficient mining of frequent patterns over data streams is proposed. This method provides a dynamic layout of sliding window by utilizing a set of simple lists for items existing within the window. For every item within the window, the most memory efficient list type based on its frequency is selected to store its occurrence information. A novel window adjustment technique including list type conversions is used to control the memory usage when the concept change occurs. At any time, if a user issues a request for frequent patterns in the recent window, a suitable approach based on the current content of the window is selected for the mining process. In comparison with recently proposed algorithms, empirical results show the superiority of the proposed method with multiple orders of magnitude in terms of runtime and memory usage.