Good afternoon everyone,
I know it seems a very straightforward question,
but why in the model evaluation assignment we used the results of the validation dataset, are not we supposed to calculate those metrics based on the test dataset predictions of the mode, or we calculate it for each one ( train_df , valid_df , test_df)
The test datasets are used to examine that whether the model works or not. And the validation datasets are used to tune the hyperparameters in the model or the other way to optimize the model.
Hello @me_sajied,
Thanks for your reply.
So which dataset we are supposed to use to evaluate the model _ calculate the accuracy, specificity, sensitivity, ppv, npv, AUCROC, … ETC
is it the validation or the test?
It is training data, precision-recall and validation dataset
You can refer to this article for more clarity.