Azure AI endpoint for the Bitcoin price prediction gives poor quality forecast

[aferdoash]

What happened + What you expected to happen

When reproducing the Bitcoin Price Prediction use case mentioned on the documentation page, it was observed that the forecasts produced are of poorer quality than displayed on the doc page.

I am however able to reproduce the forecast results shown on the doc page via the Nixtla native model.

What is the reason for the difference between these 2 models? Also, why are the data requirements different (eg. min data requirement)? Is there a way to modify this?

Versions / Dependencies

Click to expand

Dependencies: Azure AI endpoint

Reproducible example

# paste your code here
forecast = nixtla_client.forecast(df, h=7, level=level, add_history=True,model='azureai')
nixtla_client.plot(df, forecast, level=level)

Issue Severity

Medium: It is a significant difficulty but I can work around it.

[elephaint]

Hi, thanks for reporting. Is the only difference in your code the model you chose?

[aferdoash]

yes, everything else is the same