@Gina DiVenti

Thanks for your patience and the quick reply. I wasn’t able to reproduce the issue on my side, which makes me think there might be some underlying aggregations or other factors affecting the calculation.

To move forward, we have a couple of options:

• 1. If confidentiality isn’t a concern, you can share the full dataset with us and we’ll try to reproduce the issue.

• 2. Alternatively, you can raise a bug and share the details there. The more information you can include, the easier it will be for us to investigate.
  
  Below are my observations:
  The predictions when 3 departments are part of the table.

I have separated Department 1 Data and stored it as a separate file. Below are the predictions when we have only department 1 data.

I see exactly same values when I use the same values.

Explanation for Lower Values in Initial Forecasts:

The lower values in the initial forecast occur because an ARIMA model is being used. In ARIMA, the autoregressive (AR) component places greater weight on recent observations, while the moving average (MA) component smooths the series by incorporating past forecast errors. Recent historical values are in the mid-30s, the model anchors the early forecast periods around these levels. As a result, the projected values in the immediate future appear slightly lower before gradually adjusting over time.

In scenarios where strong trend or seasonality effects are present, a Prophet model is often more suitable, as it explicitly models trend changes and seasonality and can adapt more quickly to such patterns.

@Gina DiVenti

FDI data has aggregations and slightly more complex scenarios which may have influenced. At this point I can only speculate. Please feel free to raise a bug and explain the conditions. Make sure add as many details about build, versions, environment etc to narrow down. Have a great day.