The naive forecast is presented with a forecast horizon (window ahead) of one, and that makes it look very good. Its limitations would be more visible with a forecast horizon of 10 or 100… Generally, I think the forecast horizon is missing in the course (I am now in Week 4)
1 Like
Good point!
As you mentioned the prediction horizon is crucial. Feel free to use a benchmark for comparison reasons (e.g. naive prediction, assuming a constant behaviour or linear trend) when evaluating your prediction model.
Here some illustration:
I also can echo that the forecast horizon should be chosen in a meaningful way, dependent on the business problem.
Do you have any open question here or need further support?
1 Like
Thanks for your quick response! No I don’t have any open questions. Please echo that I think the course would have benefited with working with a bigger forecast horizon. The naive forecast could be one of the ones here (as you also suggest) 3.1 Some simple forecasting methods | Forecasting: Principles and Practice (2nd ed).
Anyway. I am happy with the course.
Have a nice day!