cheatsheet

Scikit-learn Cheatsheet: Inspection (Interpretability)

The inspection module provides tools to understand how a model makes decisions and which features are most important.

What can be done?

• Feature Importance: Quantify how much each feature contributes to the prediction.
• Partial Dependence: Visualize the relationship between a target and a feature, averaging out others.
• Decision Boundaries: Plot how a classifier separates classes in a 2D plane.

Key Tools

1. permutation_importance:
   • Measures importance by randomly shuffling a single feature and seeing how much the score drops.
   • Advantage: Model-agnostic and doesn’t overemphasize high-cardinality features (unlike Random Forest’s impurity-based importance).
2. PartialDependenceDisplay:
   • Shows the marginal effect one or two features have on the predicted outcome.
   • Helps identify if the relationship is linear, monotonic, or complex.
3. DecisionBoundaryDisplay:
   • Convenient tool to plot the class decision boundaries in a 2D feature space.

Comparison: Global vs Local Importance

• Impurity-based (RF): Fast, but biased toward features with many unique values.
• Permutation-based: Slower, but more reliable and works on any model.

Code Snippet: Permutation Importance & PDP

from sklearn.inspection import permutation_importance, PartialDependenceDisplay
from sklearn.ensemble import RandomForestClassifier

clf = RandomForestClassifier().fit(X_train, y_train)

# 1. Permutation Importance
# n_repeats: number of times to shuffle each feature
result = permutation_importance(clf, X_test, y_test, n_repeats=10, random_state=42)
# result.importances_mean contains the importance scores

# 2. Partial Dependence Plot (PDP)
# features: index or name of features to plot
display = PartialDependenceDisplay.from_estimator(clf, X, features=[0, 1, (0, 1)])
# (0, 1) plots a 2D interaction contour

Code Snippet: Decision Boundary Display

from sklearn.inspection import DecisionBoundaryDisplay
import matplotlib.pyplot as plt

# Only works with 2 features (X has shape [n, 2])
disp = DecisionBoundaryDisplay.from_estimator(
    clf, X, response_method="predict", cmap=plt.cm.RdYlBu, alpha=0.8
)
plt.scatter(X[:, 0], X[:, 1], c=y, edgecolors="k")

---

Credits: This cheatsheet is based on the scikit-learn documentation and examples, which are licensed under the BSD 3-Clause License. Copyright (c) 2007 - 2026 The scikit-learn developers. All rights reserved.

This site is open source. Improve this page.