Decision tree pruning
term_id: decision_tree_pruning
Category: training_techniques
Definition
Pruning is a method used to prevent overfitting in decision tree models by removing branches that have weak predictive power. It can be performed pre-pruning, by stopping the tree growth early, or post-pruning, by removing nodes from a fully grown tree. By simplifying the model, pruning improves generalization performance on unseen data and reduces computational cost during inference, making the model more robust and efficient.
Summary
A technique to reduce the size of decision trees by removing sections that provide little power to classify instances.
Key Concepts
- Overfitting prevention
- Pre-pruning
- Post-pruning
- Model complexity
Use Cases
- Improving model generalization
- Reducing inference latency
- Simplifying rule extraction