Model Interpretability, Explainability, and Feature Importance with XGBoost
While building predictive models, accuracy is often the primary focus. Model interpretability and explainability are good features to have; they are essential for trust, validation, regulatory compliance, and debugging. As machine learning models – particularly ensemble methods such as XGBoost – become more complex, understanding how these models make decisions may not be immediately apparent.
In this chapter, you’ll explore the importance of model interpretability and explainability with XGBoost, and practice extracting feature importance.
We’ll cover the following topics:
- Why interpretability and explainability matter
- Implementing XGBoost’s feature importance
- Exploring SHAP for model interpretation
- Implementing LIME for model interpretation
- Applying ELI5 for model interpretation
- Exploring PDPs for model interpretation
