Are decision trees trying to maximize information gain or entropy?

Viewed 6501

I understand that decision trees try to put classifiers with high entropy high on the decision tree. However, how does information gain play into this?

Information gain is defined as:

InformationGain = EntropyBefore - EntropyAfter

Does a decision tree try to place classifiers with low information gain at the top of the tree? So entropy is always maximized and information gain is always minimized?

Sorry, I am just a bit confused. Thank you!

2 Answers
Related