What is the role of entropy vs Gini in model selection?

Entropy and Gini are impurity measures; Gini is faster while entropy is more theoretically grounded but often yields similar trees.

Entropy measures information uncertainty using logarithmic probability, while Gini measures misclassification probability. In practice, both produce very similar splits because they are monotonically related. Gini is computationally cheaper because it avoids logarithms, making it preferred in many implementations like CART. Entropy is often used in theoretical contexts and ID3/C4.5 algorithms. The choice rarely affects final model performance significantly.