Date Added: Jun 2009
The authors want to emphasize that it is difficult to extend the discord definition from that targeted at static time series to trajectory stream scenario. That is the major reason why the authors have this paper. This paper is not about "Efficient anomaly mining in static time series". Discord definition is very useful for static time series. Although it has the defect as they pointed out (missing very similar and co-occurred discords), this defect is also considered minor by them. It is not the minor defect that leads to this work, but the fact that discords are difficult to apply for trajectory stream scenario. The authors' major contribution is to define and monitor anomalies on trajectory streams, rather than demonstrate their anomaly definition is better than discord for static time series.