Maintaining the Maximum Normalized Mean and Applications in Data Stream Mining

摘要

In data stream mining many algorithms are based on fixed size sliding windows to cope with the aging of data. This despite of some flaws of fixed size windows. Namely, it is difficult to set the size of the window and there does not exist an optimal window size due to different types of changes in the underlying distribution of the data stream. Because of these reasons the algorithm performance degrades. We propose some initial steps toward efficiently equipping sliding window algorithms with flexible windowing. This is done by the efficient maintenance of a statistic called, the maximum normalized mean. This statistic is maximized over all time windows and thus uses flexible windowing. We show that several algorithms can be restated such that it uses the maximum normalized mean as a building block. The usefulness of the normalized mean in the context of these algorithms is shown by means of experiments.