ABSTRACT

In this chapter, the authors focus on partial-dependence (PD) plots, sometimes also called PD profiles. Comparison of sub-group-specific profiles may provide important insight into, for instance, the stability of the model’s predictions. PD profiles are also useful for comparisons of different models: Agreement between profiles for different models is reassuring, Disagreement between profiles may suggest a way to improve a model and Evaluation of model performance at boundaries. Comparison of clustered or grouped PD profiles for a single model may provide important insight into, for instance, the stability of the model’s predictions. PD profiles can also be compared between different models. One of the main challenges in predictive modelling is to avoid overfitting. The issue is particularly important for flexible models, such as random forest models. PD profiles, presented in the chapter, offer simple way to summarize the effect of particular explanatory variable on the dependent variable.