Dear all,
One of the key points in TSAD (Time Series Anomaly Detection) when using a sliding window is the size of the window. For (quasi)-periodic time series, methods such as AutoPeriod, Autocorrelation etc. can be used.
But for non-periodic time series, what do you think are the best methods to use (with bibliographic references please)?
thanks :-)
What type of time series do you have in mind? For monthly or quarterly data, I can give references and suggestions for software solutions. There are even recent works on other types of data, such as daily observations.
Dear Guy
The question is a litle bit 'general'. I would like to consider that I do not have the information you mention. For example imagine data from an IOT (or ECG) :the data are 'online" without the notion of "month" or "quartely". However if you have ref in this context (yours or mine) I'm interested in :-)
I have a few references in the enclosed 2014 poster. See on p. 3, section "Outliers". I have never investigated the kind of data you mentioned.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication