Outliers naming

[axellpadilla]

First time checking this tool, looks amazing!

One question, outliers is a very general term, and is widely used in the industry for samples in time series too, while I understand anomaly naming if it is going to be included later, maybe the naming could be confusing, ts_outliers for example is a very used term, is there going to be a clear way to differentiate by the IDE suggestions for example, if one thing is for ts or not?

[sd2k]

Hey @axellpadilla. Yeah we run into this all the time with our ML features at Grafana - the terms 'anomaly' and 'outlier' are used almost interchangeably within the industry. We basically drew the line similarly to Datadog I think:

• anomalies are within time series: individual samples behaving differently to expected
• outliers are between time series: one or more time series behaving differently to the rest.

I think we could make that clearer within the docs but I don't want to have to add it to every doc comment ideally, it'll get boring fast! Do you have any suggestions for where the docs are lacking?

[axellpadilla]

Taking into account that is indeed an accepted way to call both the same, maybe you could include both functionalities on the same outliers, receive just 1 series, do it in series, receive more do series vs series, that will keep all the functionality easily discoverable, but I really don't know if that will be appropiate. You could also have just a different namespace for individual time series analysis and set of ts analysis, and or, add aliases like ts_outliers, ts_anomalies and set_outliers, but I'm personally more inclined to just accept we call outliers everything haha.

I'm looking for the python usage examples but can't find any, I see there are not changepoint or outliers detection bindings, is that right?

And also, congrats, your demo is impressive!