Skip to content
ML Academy

Introduced on day 23

Information criterion (AIC, AICc)

A score that adds a penalty for the number of fitted parameters to a model's in-sample negative log-likelihood; it ranks models of one class on the same data and does not measure forecast error.

In detail

AICc adds a small-sample correction to AIC. Because the likelihood is computed differently for ETS and for ARIMA (which is fitted on differenced data), the values cannot be compared across those classes; within a class, models must be fitted to the same series, with the same transformation and differencing. Out-of-sample accuracy needs forecast errors on held-out data.