DataTerminology.comNEW
Also known as: MAPE, Mean Absolute Percentage Deviation
Average of absolute percentage errors between predictions and actual values.
DESCRIPTION
Mean Absolute Percentage Error expresses prediction errors as percentages of actual values. It is intuitive for stakeholders but problematic when actuals are near zero and can overweight small denominators. MAPE is scale-independent, aiding cross-series comparisons. Variants like sMAPE and MASE address some shortcomings. Use MAPE carefully with safeguards for zeros and robust evaluation protocols.
ALTERNATIVE DEFINITIONS
Wikipedia
The mean absolute percentage error (MAPE), also known as mean absolute percentage deviation (MAPD), is a measure of prediction accuracy of a forecasting method.
Frames MAPE as a forecasting accuracy measure and notes alternate name.
Forecasting: Principles and Practice
The mean absolute percentage error (MAPE) is the mean of the absolute percentage errors.
Concise textbook definition used widely in forecasting.
RELATED TERMS
RMSE Root Mean Squared Logarithmic Error Forecast Accuracy Systematic Error Outliers Zero Values Regression Metrics Forecasting
COMPETING TERMS
Mean Absolute Error Symmetric Mean Absolute Percentage Error
KNOWLEDGE AREAS
Data Science AI Mathematics Testing & Validation
TARGET AUDIENCE
Data Scientist Data Analyst AI Engineer MLOps Engineer
EXTERNAL SOURCES
- Wikipedia - Mean absolute percentage error
- scikit-learn - MAPE definition and caveats in sklearn.
- Hyndman & Athanasopoulos - Forecast accuracy measures including MAPE.
TOP QUESTIONS
What is Mean Absolute Percentage Error?
MAPE is the average absolute percentage difference between predicted and actual values. It is intuitive and scale-free.
Category: Definition
When should I use MAPE?
Use MAPE to compare forecast accuracy across series with different scales, provided actuals are not near zero.
Category: Relevancy
What are MAPE’s limitations?
It can blow up near zero actuals and tends to overweight small denominators. It is asymmetric and biased toward over-forecasts.
Category: Risk
How can I mitigate MAPE pitfalls?
Filter or adjust zero values, consider sMAPE or MASE, and report multiple metrics for robustness.
Category: How-To
How is MAPE computed?
It is the mean of |(y − ŷ)/y| expressed as a percentage, often multiplied by 100.
Category: Requirements
What is a common misunderstanding about MAPE?
Assuming it is always comparable across datasets. Distributional differences and zeros can distort comparisons.
Category: Misunderstanding
What are alternatives to MAPE?
Consider MAE, RMSE, sMAPE, MASE, or RMSLE depending on scale, zeros, and business context.
Category: Alternatives
Is MAPE robust to outliers?
It is less sensitive than RMSE but can still be skewed by small actuals or extreme relative errors.
Category: Opinion
How to report MAPE to stakeholders?
Present it with context, confidence intervals, and complementary metrics to avoid misleading conclusions.
Category: Communication