Computer scienceData scienceMachine learningRegressionRegression metrics

MAPE and sMAPE

The scale-independence

Report a typo

We mentioned how MAPE and sMAPE are scale-independent, meaning that the error scores are not affected by the scale of the underlying data.

Which factor allows us to make comparisons between different models or datasets?

Select one option from the list

Lower scores indicate better model performance

MAPE and sMAPE treat negative and positive errors unequally

Errors are expressed in percentages (and thus, have a more standard range)

Both sMAPE and MAPE are based on taking absolute errors

___

Create a free account to access the full topic