is it usefull to retrain the selected modell on the full dataset after tested on split data?

brys

i have a basic data-science theoretical question (apoligies if answered somewhere, i could not find it): after selecting a satisfying modell (using split datasets via k-fold cross-validation) is it usefull to retrain the modell on the full dataset before using it for prediction?

MartinLiebig

Hi,

the model you get from cross validation is exactly doing this, because it is very likely that the model is better.

Best,

Martin