Decision Trees in Machine Learning (2026 Guide)

A classification tree predicts a category (and often a probability per class). A regression tree predicts a number, typically using the average target value within each leaf.

At each node, the tree tests possible features and thresholds and chooses the split that most improves its objective, such as reducing impurity for classification or reducing error for regression.

Trees can keep splitting until they capture very small patterns in the training data. Without constraints, they can memorise noise, leading to strong training results but weaker performance on new data.