Implementation:Interpretml Interpret PartialDependence

Metadata

Field	Value
Sources	Repo: InterpretML
Domains	Interpretability, Feature_Analysis
Updated	2026-02-07
Type	API Doc

Overview

Concrete tool for computing partial dependence plots for blackbox models provided by the InterpretML library.

Description

The PartialDependence class computes PDP data during __init__ (not during explain_global). It creates a grid of values for each feature, evaluates the model at each grid point (averaging over samples), and stores the results. The explain_global() method formats these precomputed results into an Explanation object for visualization.

Usage

Use this when you want to understand global feature effects for any model through the InterpretML show() pipeline.

Code Reference

Field	Value
Source	interpretml/interpret
File	python/interpret-core/interpret/blackbox/_partialdependence.py
Lines	72-191
Import	`from interpret.blackbox import PartialDependence`

Signature:

class PartialDependence(ExplainerMixin):
    available_explanations = ["global"]
    explainer_type = "blackbox"

    def __init__(self, model, data, feature_names=None, feature_types=None, num_points=10, std_coef=1.0):
    def explain_global(self, name=None):

I/O Contract

Init inputs:

Parameter	Type	Required	Default
model	predict function	Yes
data	ndarray	Yes
feature_names	list	No
feature_types	list	No
num_points	int	No	10
std_coef	float	No	1.0

explain_global inputs:

Parameter	Type	Required	Notes
name	str	No

explain_global output: PDPExplanation with partial dependence curves per feature.

Usage Examples

from interpret.blackbox import PartialDependence
from interpret import show

pdp = PartialDependence(rf.predict_proba, X_train, feature_names=feature_names)
pdp_global = pdp.explain_global(name="Partial Dependence")
show(pdp_global)
show(pdp_global, key=0)  # First feature's PDP

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment