1
$\begingroup$

Suppose we stop growing the tree here at gender enter image description here

Then if a new sample has moderate education will consider the label to be true or false? Looks like it will be false 66% of the time but this isn't very helpful.

$\endgroup$

1 Answer 1

2
$\begingroup$

In the decision tree, no matter if you prune it or not, the decision is made by a majority vote. If I see correctly, 2/3 of the samples are marked "No", so the decision is "No". Usually, you would deal with large samples of data and you would have some criteria on the depth of the tree or size of the nodes that, same as pruning, would lead to end nodes that have samples of more than one class, so you would do majority vote. As you correctly noticed, 2/3 is also an estimate for the probability predicted by the decision tree, though such probabilities are not well calibrated and usually not that useful in practice.

$\endgroup$

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.