Akaike Information Criterion

MuehliMan

Hi all,

if it possible to calculate the Akaike Information Criterion or even better the Schwarz Infoprmation Criterion by some Performance Operators?
I would like to use it for the Optimize Selection subprocess.

Best,
Markus

land

Hi,
the AIC isn't available in RapidMiner and that has a reason: AIC is only a heuristic to estimate a test error based upon the training error. It is much more valid to estimate this test error by applying a cross-validation.
If you want to add such an criterion, how would you calculate the number of parameters a model like the SVM or KNN has? I don't know if this is always determined.

Greetings,
  Sebastian