Computer scienceData scienceMachine learningRegressionRegression metrics

MAPE and sMAPE

Close to zero

Report a typo

Both MAPE and sMAPE are know to exhibit specific behavior when dealing with values close to zero. Select the applicable statements from the list below.

Select one or more options from the list

MAPE is well-suited for the cases when the true target values might be equal or close to zero

sMAPE is well-suited for the cases when the true target values might be equal or close to zero

In the case of sMAPE, if either the prediction or the ground truth is equal to zero, the score will be maximized

Values close to zero cause MAPE to lack the upper bound (in other words, they tend to infinity)

___

Create a free account to access the full topic