Mean Absolute Error (MAE)

This value represents the mean of absolute differences between actual values and forecasts for each unit of time. MAE is an error metric that is easy to understand and also robust to outliers. A lower value of MAE implies a higher accuracy of the model. MAE expresses the average forecast error in units of the variable of interest, and as a result it cannot be used to make comparisons between time series that involve different units. A forecast method that minimizes the mean absolute error will lead to forecasts of the median.