What is a binary split in Decision Trees?
Updated May 16, 2026
Short answer
A binary split divides a node into exactly two child nodes based on a threshold condition.
Deep explanation
In decision trees, a binary split evaluates a condition like feature <= threshold and sends data into left or right branches. Most modern implementations like CART (Classification and Regression Trees) strictly use binary splits because they simplify optimization and improve computational efficiency. The algorithm searches for the best threshold that minimizes impurity (Gini, entropy, or MSE). Even categorical features are often converted into binary partitions internally.
Real-world example
In credit scoring, a binary split might check if income <= 50,000 to separate low-risk vs high-risk applicants.
Common mistakes
- Assuming decision trees always split into multiple branches per node instead of binary splits in CART.
Follow-up questions
- Why do CART models prefer binary splits?
- Can decision trees handle categorical variables directly?
- How is the best threshold chosen?