This paper addresses the problem of similarity detection in time series from both an effectiveness and efficiency viewpoint. A main motivation underlying our work is that, as we experimentally proved, no state-of-the-art method has capabilities for both fast and accurate similarity detection in time series. Viewed in this respect, we propose DSA (Derivative time series Segment Approximation), a representation model for time series that suitably combines the notions of derivative estimation, segmentation and segment approximation to support effective and efficient similarity detection. Experiments conducted in a hierarchical clustering framework show that DSA based similarity detection is as good or better than both the most accurate and the fastest methods among the competing ones.
Accurate and fast similarity detection in time series
TAGARELLI, Andrea;GRECO, Sergio
2007-01-01
Abstract
This paper addresses the problem of similarity detection in time series from both an effectiveness and efficiency viewpoint. A main motivation underlying our work is that, as we experimentally proved, no state-of-the-art method has capabilities for both fast and accurate similarity detection in time series. Viewed in this respect, we propose DSA (Derivative time series Segment Approximation), a representation model for time series that suitably combines the notions of derivative estimation, segmentation and segment approximation to support effective and efficient similarity detection. Experiments conducted in a hierarchical clustering framework show that DSA based similarity detection is as good or better than both the most accurate and the fastest methods among the competing ones.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.