Hello! I am just trying out TimeGPT for the first time. I see in the documentation that the example uses just a time series column and a target column, is there a way to provide my own features to the model, and additionally tune their weight/importance relative to the other data TimeGPT is using? If there is more in depth documentation on TimeGPT I'm sure I can find what I am looking for there. Thank you!

Hello! Yes you can pass exogenous variables or automatically generate some features using the

date_features

parameter. You can see a complete example on using TimeGPT with exogenous variables here. It also shows how to extract the weights of the features. I hope this helps!

@Marco Cool, thank you for this! I am currently toying with anomaly detection with exogenous variables. I am getting strange behavior from the model and was just wondering if you could shed some insight. I will attach my df, my anomaly model run line, and the resultant anomalies. I was able to run this successfully using a different feature as the target, but for some reason this specific feature is giving me issues

Hey @Kyler Suden. This was an issue on our side, can you please try again?

Sure, give me a few moments

Sorry it seems the deployment wasn't updated correctly. Can you try again?

Yes that worked, thank you guys for looking into this!

Thanks for the detailed report!