interpretml/interpret

Fit interpretable models. Explain blackbox machine learning.

What it solves

InterpretML provides a unified framework to make machine learning models more transparent. It addresses the "black box" problem where complex models make predictions without clear reasons, which is critical for debugging, feature engineering, detecting bias, ensuring regulatory compliance, and deploying AI in high-risk fields like healthcare and finance.

How it works

The package offers two primary paths to interpretability:

  1. Glassbox Models: These are inherently interpretable models. The standout is the Explainable Boosting Machine (EBM), which uses bagging and gradient boosting to achieve accuracy comparable to random forests while remaining fully transparent and editable by experts.
  2. Blackbox Explainers: For models that are already complex (like deep neural networks), it provides tools like SHAP, LIME, and Partial Dependence to explain why a specific prediction was made or how the model behaves globally.

Who it’s for

Data scientists and ML engineers who need to understand, trust, and audit their models, as well as domain experts who need to edit model behavior based on their professional knowledge.

Highlights

  • Explainable Boosting Machine (EBM): A high-performance glassbox model that matches the accuracy of state-of-the-art blackbox models.
  • Unified Interface: Combines multiple techniques (APLR, Decision Trees, SHAP, LIME) under one roof.
  • Privacy-Preserving: Includes Differentially Private EBMs to protect sensitive training data.
  • Causal Analysis: Supports global and local explanations to understand both overall model behavior and individual predictions.

Related

  • Project
    Trusted-AI/AIX360

    An open-source library providing a comprehensive set of algorithms and metrics to improve the interpretability and explainability of machine learning models and datasets across various data modalities.

  • Project
    mljar/mljar-supervised

    An Automated Machine Learning package for tabular data that automates preprocessing, model selection, and tuning while providing transparent, detailed reports for explainability.

  • Project
    openml/automlbenchmark

    A framework for evaluating and comparing open-source AutoML systems using standardized environments and curated datasets to ensure reproducible research.

  • Project
    MAIF/shapash

    A Python library that makes machine learning models interpretable for everyone by providing user-friendly visualizations, an interactive web app, and audit reports.

  • Project
    microsoft/responsible-ai-toolbox

    A suite of model and data exploration tools that enable developers to identify errors, assess fairness, and interpret AI model predictions for more responsible deployment.