Time Series values differ from original data
Given an Original Data set, then do Time Series Lab, and deploy a model.
Then, I train the model > predict > score based on Original Dataset, getting Forecast Data.
Forecast data contains date, values, forecast, percentile columns. I assume values to be the original data. However, when I compare it to my original data it is not the case. The differences are generally (but not always) small, but also not small enough to be precision errors, and large enough to be noticeable. Am I mistaken in my understanding of what Value column is? Why does this happen?
Bonus Question:
Original Data is 'start of months' , like Jan'24 is 2024-01-01,
After forecasts data becomes 'end of months', Jan'24 becomes 2024-01-31
I would like to keep it as start of month but do not have the lates time series version due to company restrictions. Is it possible to keep the formatting? Also I am then curious to know if each record becomes a forecast for month up to this date, or month from this date.
ie. is 2024-01-31 a forecast for January forecast, or February?