Chapters

This chapter introduces the basic concepts of Interpretable Machine Learning. We focus on supervised learning, explain the different types of explanations, repeat the topics correlation and interaction.

Some machine learning models are already inherently interpretable, e.g. simple LMs, GLMs, GAMs and rule-based models. These models are briefly summarized and their interpretation clarified.

Feature Effects indicate the change in prediction due to changes in feature values. This chapter explains the feature effects methods ICE curves, PDP and ALE plots.

Shapley values originate from classical game theory and aim to fairly devide a payout between players. In this section a brief explanation of Shapley values in game theory is given, followed by an adaption to IML resulting in the method SHAP.

Feature importance (FI) methods rank features by how much they contribute to the predictive performance or prediction variance of the model. With a loss-based focus this section deals with the FI methods PFI, CFI, SAGE and LOCO.

A common approach to interpret an ML model locally is implemented by LIME. The basic idea is to fit a surrogate model while focussing on data points near the observation of interest. The resulting model should be an inherently interpretable one.

This chapter deals with further local analyses. First, counterfactuals are examined which search for data points in the neighborhood of an observation that lead to a different prediction. Second, the robustness of an ML model is checked by explicitely searching for malicious inputs, which are called adversarial examples.