Identification of outliers in simultaneous time series using Wavelets and Forward Search
In this paper we present an original methodology for the processing of simultaneous time series. It allows to identify times at which two time series present different patterns - these being outliers in a certain sense. We first apply wavelet techniques to transform data into representations focusing on singular locations. Several approaches are proposed to obtain such representations of the data. Then we explore these with a modern tool in robust statistics, the Forward Search, which allows in particular to evidence outliers.
DAMERVAL Christophe;
2013-01-25
Publications Office of the European Union
JRC72903
978-92-79-26269-2,
1831-9424 (online),
EUR 25493 EN,
OP LB-NA-25493-EN-N (online),
https://publications.jrc.ec.europa.eu/repository/handle/JRC72903,
10.2788/47188,
Additional supporting files
| File name | Description | File type | |
|---|
