Retrain model on the whole data set (include test set) when deploying a model

Hi, after I got the best hyper parameter that give me the highest performance on the test set. If i am going to deploy the model, should I retrain (or refitting) the model with the same best hyper parameter on the whole dataset again (including test set) or not. Because as far I know it would be good if the model see lot of data as much as possible. Or we just use the model that training on the cross validation when implement.

1 Like

No, do not re-train after you verify the performance with the test set.

The test set results are your proof that the model works well enough. There is no need to train again after that.

You don’t need as much data as possible. You need enough data to get good-enough results.

No, that’s not it either. You train on the training set. You don’t train on the validation set.