Naive forecast baseline
A simple forecast rule without training, for example “as in the last period”; a complex method or an AI is compared with it.
Formula
Improvement = 1 − the method’s error ÷ the baseline error
How to read
A method that does not beat the baseline is not needed however it looks.