MAPE is a metric for evaluating forecasting or regression models. It measures the average absolute error between predicted and actual values, expressed as a percentage of the actual values. This makes it intuitive for communicating model accuracy to stakeholders or management. Lower values indicate better model performance.
Formula:
Notes:
- Useful for comparing errors across different scales.
- Sensitive to very small actual values, which can inflate the percentage error.
Related:
MAPE vs RMSE
For an initial forecast model, use MAPE over RMSE because it provides a relative measure of error that’s easy to interpret. By expressing errors as a percentage of actual values, MAPE is scale-independent and gives a intuitive sense of how far off forecasts are.
MAPE also treats all periods more evenly, while Root Mean Squared Error gives extra weight to large errors, which can skew evaluation if outliers or volatile periods exist.
MAPE provides a simple, easy-to-read overview of forecast accuracy on test data and helps guide improvements before moving on to more sensitive metrics like RMSE or MAE.