How to choose model to deploy

Let’s say, I have done my cross validation for two model which is linear regression and random forest when choosing the model to deploy in production should i choose it from k fold cross validation score or should I use that both model to test it on test set and choose the best one the have best performance on test set. My first understanding is after we get some best model from cross validation we choose it again with test set. But i saw many kaggle competition the most or the winner choose the model from cross validation score.

In my experience K-fold validation is often used if the data set is too small to split into separate training, validation, and test sets that are large enough to be statistically significant.

Remember that Kaggle hosts contests where they need a well-defined numerical ranking among the competitors. It has different goals than you would face in real-world solutions.

Hello @peachans,

As explained in the course, we use the cross validation score to pick the best model.