Train, Test and CV set on decision trees

Why is it that I am not using both a test and CV set while evaluating the accuracy for decision trees? Since the parameters are being adjusted on the calibration set, aren’t those now considered model features? Wouldn’t it be necessary a CV set to evaluate accuracy without bias?

If you perform cross validation then I guess there would be no need to have a test set, otherwise its advisable to have a test set.

In W4_Lab_01 we build decision tree and at the end we can see the tree itself. How can I afterwards use my test sample and use it on trained model to get and inference?

Sorry, I don’t have access to this course but surely they will do the inference and testing as labs progress!

1 Like