Principle:Interpretml Interpret Partial Dependence Explanation
Metadata
| Field | Value |
|---|---|
| Sources | Paper: Friedman 2001 "Greedy function approximation: a gradient boosting machine", Doc: sklearn PDP |
| Domains | Interpretability, Feature_Analysis |
| Updated | 2026-02-07 |
Overview
A global explanation method that shows the marginal effect of one or two features on a model's predicted outcome by averaging over the other features.
Description
Partial Dependence Plots (PDP) visualize how a feature affects predictions on average, marginalizing over all other features. For each value of the feature of interest, the model is evaluated on all samples with that feature set to the grid value, and the results are averaged. The resulting curve shows the average relationship between the feature and the prediction.
Usage
Use PDP when you want to understand the average effect of individual features on any black-box model's predictions. It provides global insight complementary to local methods like SHAP or LIME.
Theoretical Basis
The partial dependence function for feature set S is:
Where S is the feature of interest and C is the complement set.