Mean Absolute Error (MAE) is used to evaluate regression problems. MAE is the mean of the total absolute difference between the true and predicted value. Large and small errors are treated equally important in MAE since the values are not squared as in Mean Squared Error.