How do Decision Trees approximate nonlinear decision boundaries?

Decision Trees approximate nonlinear boundaries using a sequence of axis-aligned splits.

Although each split is a simple threshold on a single feature, combining many splits allows decision trees to approximate complex nonlinear boundaries. Each split partitions the feature space into rectangular regions. As depth increases, these regions become smaller and more complex, allowing the tree to approximate curved or nonlinear decision surfaces. However, this requires deeper trees, which increases variance and risk of overfitting.