interpret-core

What it is and what it does

interpret-core is a unified framework for training interpretable models and explaining blackbox machine learning systems. It implements state-of-the-art techniques including Explainable Boosting Machines (EBMs)—a modern take on Generalized Additive Models that achieve accuracy comparable to random forests and gradient boosting while remaining fully interpretable—alongside explainers for existing models like SHAP, LIME, and partial dependence analysis. The package is built on numpy, pandas, scikit-learn, and joblib, making it a natural fit for standard data science workflows.

interpret-core addresses critical use cases in model debugging, feature engineering, fairness detection, regulatory compliance, and high-risk domains like healthcare and finance. You can train glassbox models directly, explain individual predictions, compare multiple models side-by-side, and optionally apply differential privacy for sensitive data. The framework handles pandas DataFrames and numpy arrays natively and includes support for pairwise interactions by default.

Use it for:

• Train an Explainable Boosting Machine for classification or regression when you need both accuracy and full model transparency.
• Debug why a blackbox model made a specific prediction using SHAP or LIME explainers on existing models.
• Detect fairness issues by analyzing global feature importance and local prediction explanations across demographic groups.
• Satisfy regulatory requirements in finance or healthcare by producing exact, auditable explanations for individual decisions.
• Engineer better features by understanding which variables drive model predictions and how they interact.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Trains interpretable machine learning models and explains blackbox predictions using techniques like Explainable Boosting Machines, SHAP, LIME, and decision trees.

Yes. interpret-core is actively maintained, has low install friction, and fills a genuine need for interpretable ML in regulated and high-stakes domains. The license treatment is unclear—verify the actual license before use in proprietary contexts—but the package itself is stable and has no known vulnerabilities. Install if you need interpretability alongside accuracy.

Install

interpret-core on PyPI

pip install interpret-core

uv add interpret-core

poetry add interpret-core

Installing interpret-core

Before you install

Low friction installation with a pure Python wheel. Active maintenance with recent releases and solid community engagement (6916 stars). Four runtime dependencies—numpy, pandas, scikit-learn, joblib—are standard data science libraries.

License in practice

License treatment is unclear; the package metadata does not specify an SPDX identifier or raw license string. Verify the actual license terms in the repository before adopting in proprietary or regulated contexts.

Quickstart

pip install interpret-core

from interpret.glassbox import ExplainableBoostingClassifier

ebm = ExplainableBoostingClassifier()
ebm.fit(X_train, y_train)

from interpret import show
show(ebm.explain_global())

Requires Python 3.10 or later.

Verify before relying

• Whether the unclear license permits use in proprietary or commercial applications.
• Performance characteristics on datasets larger than typical in-memory workloads.
• Compatibility with GPU acceleration or distributed training frameworks beyond Azure SynapseML.

Package facts

Evidence: interpret_core-0.7.8-py3-none-any.whl

Tags

More Artificial Intelligence packages