Information Gain

Back to Decision Trees

The reduction in entropy (uncertainty) after splitting on a feature. Used as a criterion for choosing the best split in decision trees. Higher information gain means the feature better separates the classes.

ml decision-trees information-theory