Mean Absolute Error (MAE) is a statistical measure that evaluates the accuracy of a predictive or forecasting model by calculating the average of the absolute differences between predicted and actual values. MAE is expressed in the same units as the data and provides a straightforward interpretation of the average size of prediction errors. Analysts use the mean absolute error widely in forecasting, regression analysis, and machine learning to assess how closely model predictions align with observed outcomes.
Unlike MAPE, which expresses error as a percentage of the actual value, MAE reports error in the same units as the original data, making it an absolute (rather than relative) measure of accuracy.
The formula for the Mean Absolute Error is:
MAE = (1/n) × Σ|Actual – Predicted|
This formula takes the absolute value of the difference between each actual and predicted value, then averages those differences across all observations. Because it treats all errors equally, MAE gives a balanced picture of overall model error without exaggerating the influence of outliers.
For example, suppose a model predicts daily website traffic for five days as follows:
-
Day 1: Actual = 500, Predicted = 520
-
Day 2: Actual = 450, Predicted = 460
-
Day 3: Actual = 480, Predicted = 470
-
Day 4: Actual = 510, Predicted = 500
-
Day 5: Actual = 530, Predicted = 510
The absolute errors are 20, 10, 10, 10, and 20. The MAE is:
MAE = (20 + 10 + 10 + 10 + 20) / 5 = 14
This means the model’s predictions are off by 14 visits per day on average, according to the Mean Absolute Error.
While the Mean Absolute Error resembles the Mean Absolute Deviation (MAD), the two are conceptually different. MAE compares predicted values to actual values to assess forecasting accuracy, whereas MAD measures variability in a dataset around a central value.
« Back to Glossary Index