0
$\begingroup$

Suppose we make a linear regression model on each of the 10 folds with the same number of features (say 2 for simplification) We will therefore have 10 sets of coefficients with the optimized values ​​of the parameters with each of the metrics (for example R2).

In the end what model do we retain? (the best, the average values ​​obtained for each parameter, the model that obtained the average R2?) Ouvrir dans Google Traduction

$\endgroup$
0

1 Answer 1

1
$\begingroup$

Generally, if you are using folds, the final prediction is calculated as an average of all predictions obtained using the models based on folds. Linear regression is a special case of ML methods, where averaging the predictions gives the same results as using a model of averaged parameters. So, regarding your question, in your way of using folds and models, the proper parameters of the final model are the average values ​​obtained for each parameter.

$\endgroup$
2
  • $\begingroup$ An important point is that this would not be the same with other types of models, this applies specifically to linear regression. $\endgroup$ Dec 13, 2019 at 6:31
  • $\begingroup$ Thanks you for your contribution. $\endgroup$
    – lpean64
    Dec 14, 2019 at 9:52

Not the answer you're looking for? Browse other questions tagged or ask your own question.