Interpretable Machine Learning with Python

Chapter 3: Interpretation Challenges

In this chapter, we will discuss the traditional methods used for machine learning interpretation for both regression and classification. This includes model performance evaluation methods such as RMSE, R-squared, AUC, ROC curves, and the many metrics derived from confusion matrices. We will also explore several dimensionality reduction visualization techniques that can be leveraged for interpretation purposes. We will then examine the limitations of these traditional methods and explain what exactly makes "white-box" models intrinsically interpretable and why we cannot always use white-box models. To answer this question, we'll consider the trade-off between prediction performance and model interpretability. ...