In a good model, errors are reduced, and overfitting is avoided. Indicators to assess the model itself are also known.
-
The accuracy of the model is how much the estimated value differs from the true value, usually expressed as R².
-
Recently, AIC and BIC have also been used to take generalization performance into account.
-
These two are often used to assess model fitting, but the BIC contains a sample size component. As the sample size increases, the influence of sample size becomes dominant in the BIC values.
-
-
We also use threshold-by-threshold Bayesian probabilities from a business perspective.