164 страниц. 2010 год. LAP Lambert Academic Publishing This book provides new approaches to forecast time series, based on data mining techniques. It explores several clustering algorithms and proposes methods to exploit their strengths when dealing with temporal data. This work proposes thus a brand new philosophy to forecast any kind of time series, as all introduced methodologies claim to be general- purpose. In fact, more than one hundred real-world time series have been analyzed with great accuracy.