Table 3. Explanation of fit metrics

Fit metric Description Value range Typical values for good fit
Stationary R-squared Measures how well the model fits for the stationary part of the time series 0–1 (higher is better) Close to 1 suggests a good fit for stationary data
R-squared Indicates proportion of the variance in the dependent variable that is predicTable from independent variables 0–1 (higher is better) Close to 1 for models with high explanatory power
RMSE Root mean square error, measuring the average magnitude of residuals Non-negative real numbers Close to zero indicates minimal prediction error
MAPE Mean absolute percentage error, measuring accuracy as a percentage 0.0%–100.0% Values below 10.0% are considered as good
MaxAPE Maximum absolute percentage error, shows the largest absolute percentage error 0.0%–100.0% Lower values (close to zero) are better
MAE Mean absolute error, providing the average absolute difference between predicted and observed values Non-negative real numbers Close to zero indicates lower average error
MaxAE Maximum absolute error, indicating the largest absolute error observed Non-negative real numbers Lower values (close to zero) are better
Normalized BIC Bayesian information criterion (BIC), adjusted for model comparison on a standardized scale Real numbers (lower is better) The lowest number among compared models is preferred
Data from Bindel and Seifert (2025); Hyndman and Athanasopoulos (2018).