What is Akaike information criterion (AIC)
AIC is a measure of fit which penalises the model for having more variables.
The AIC is defined as:
AIC = n.ln[SSE] / n+2k
- n is the number of cases in the model, ln is the natural log, SSE is the sum of square errors for the model, and k is the number of predictor variables
When to use Akaike information criterion (AIC)
Trusted Data advise is that you should select models after careful examination of AIC value. In general, the low the AIC model value the better the model. However the latter shouldn’t be applied to all statistical, ML or Deep Learning models because often we are interested in loss function as well.