Discovering all frequent trends in time series
Ajumobi Udechukwu, Ken E. Barker, Reda Alhajj · 2004
Several techniques have been proposed for translating and mining time series. The translation schemes are typically based on passing a window over the time series and extracting features from the data lying within the window. The use of windows in time series translation has been shown to be effective in indexing and querying similar time series. However, for applications involving the identification of frequent patterns in time series, and finding pattern associations existing in single or multiple time series, significant domain knowledge is required to effectively choose a window size. Alternatively, an expensive all-window approach may be employed.In this work we present a linear-time, domain-independent technique for translating time series and finding all frequent trends in the series without using a time window.