Closed alecalma closed 4 years ago
Hello @alecalma,
the short answer is no: gluon-ts is not suited to detect anomalous time-series in a dataset.
Surely, some components could be reused, but the general idea of gluon-ts is to be a forecasting oriented library, making predictions about the future. There are ideas to use gluon-ts to detect anomalous values by making predictions and then compare the result to the real value, but I think that's a fundamentally different problem from yours.
Hi,
I had a look at the tutorials and they are all forecasting related. I would like to know if I could use the library for anomaly detection among different time series. Let me explain better: I have a big dataset with hundreds of time series which should be all very similar to each other. I would like to run an automated tool that identifies anomalous time-series and is able to cluster them based on some similarity metric.
Would this be possible with Gluon-TS?
Thank you.