0
$\begingroup$

I want to use cross validation to validate the results of my model, However, I'm confused how to use it.

Should I split my data to 3 sets (Train, Validation,test) and validate the results on test set?

or make the validation on the whole data set?

or use validation as a training phase and test on the test set?

It is good to mention that my dataset is almost 500 records. Thanks

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Yes, you should divide your dataset in three sets(train, validate and test).

As you are having 500 records, a good division may be [250, 125, 125]. Where you are having 250 records for training, 125 validating and 125 testing.

You may find this useful.

Improve this answer
$\endgroup$
2
  • $\begingroup$ even though I dont want to tune parameter? , I only want to validate the model I'm getting after testing? $\endgroup$
    – user201561
    Commented Mar 26, 2018 at 11:03
  • $\begingroup$ As per my understanding, Validation is itself used for tuning hyper-parameters. You've written, you don't want to tune those parameters. Then why to use validation. Correct me if I'm wrong. I also want to know what you mean by "I only want to validate the model I'm getting after testing". $\endgroup$
    – Sanjay Chandlekar
    Commented Mar 26, 2018 at 11:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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