Clarify values used in the evaluation in the forecast example

[rolerik]

In the evaluation method data from the second year is transformed by the model.
Previewing the predictions I see 7 values for ForecastedRentals in the results. I assumed that is because of the horizon = 7.

What do the values represent? Is that one week after the training data, i.e forecast starts at 2012-01-01?
Or is it forecasting after the second year has been added by transforming the testdata, i.e forecast starts at 2013-01-01?

In either case it looks like the example code only looks at the first day of each prediction (ForecastedRentals[0]).
Can that really be compared with actual values from the testdata, which has period 1 day rather than 7 days?

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: 32d03435-cc6c-1123-a86c-d5c3eacc24bd
• Version Independent ID: 460afd2b-383d-7b90-6804-4b900c3bc9b2
• Content: Tutorial: Forecast bike rental demand - time series - ML.NET
• Content Source: docs/machine-learning/tutorials/time-series-demand-forecasting.md
• Product: dotnet-ml
• GitHub Login: @luisquintanilla
• Microsoft Alias: luquinta

[rolerik]

I modified the Evaluate method to do a prediction instead, similar to Forecast method. This seems more reasonable to me in order to compare with the actual results from year 2

Modified code: Program.cs

I don't understand why the predictions are different in this case, unless the Transform used in the example code actually moved the prediction point to 2013-01-01.

However, with the modified code the results from the Forecast method looks more similar to the graph in the tutorial.
Except that the lower estimates are different.
This makes me wonder if this was the way the graph was generated in the first place?

[rolerik]

Related to #27469