2012 | OriginalPaper | Buchkapitel
Size Matters: Finding the Most Informative Set of Window Lengths
verfasst von : Jefrey Lijffijt, Panagiotis Papapetrou, Kai Puolamäki
Erschienen in: Machine Learning and Knowledge Discovery in Databases
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Event sequences often contain continuous variability at different levels. In other words, their properties and characteristics change at different rates, concurrently. For example, the sales of a product may slowly become more frequent over a period of several weeks, but there may be interesting variation within a week at the same time. To provide an accurate and robust “view” of such multi-level structural behavior, one needs to determine the appropriate levels of granularity for analyzing the underlying sequence. We introduce the novel problem of finding the best set of window lengths for analyzing discrete event sequences. We define suitable criteria for choosing window lengths and propose an efficient method to solve the problem. We give examples of tasks that demonstrate the applicability of the problem and present extensive experiments on both synthetic data and real data from two domains: text and DNA. We find that the optimal sets of window lengths themselves can provide new insight into the data, e.g., the burstiness of events affects the optimal window lengths for measuring the event frequencies.