TOPICS
Search

Mean Absolute Percentage Error

The mean absolute percentage error (MAPE) of predictions y^^_i for nonzero observed values y_i is

MAPE=(100)/nsum_(i=1)^n|(y_i-y^^_i)/(y_i)|.

It expresses forecast error as a percentage, but is undefined when an observed value is zero and can be unstable when observed values are close to zero. Its use is also problematic for data whose meaningful values may have different signs.

See also

Mean Absolute Error, Mean Square Error, Root-Mean-Square Error

Explore with Wolfram|Alpha

References

Hyndman, R. J. and Koehler, A. B. "Another Look at Measures of Forecast Accuracy." Int. J. Forecasting 22, 679-688, 2006. https://doi.org/10.1016/j.ijforecast.2006.03.001.

Cite this as:

Weisstein, Eric W. "Mean Absolute Percentage Error." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/MeanAbsolutePercentageError.html

Subject classifications