InterpretML

What it does

InterpretML is a Python package that takes in tabular training data (pandas dataframes or numpy arrays) and a model, and produces explanations of how that model behaves. It can train glass-box models such as Explainable Boosting Machines, linear models and decision trees, or wrap an existing black-box pipeline and explain it with methods like LIME, SHAP, partial dependence and Morris sensitivity analysis. Output is interactive visualizations showing global feature importance, local explanations of single predictions, and model comparisons.

Practitioners typically fit a model with a scikit-learn style interface, call explain_global or explain_local, then view results with show() in a notebook. A differentially private variant of EBM is included for sensitive data, and distributed EBMs are available on Azure SynapseML for very large datasets. Its distinguishing feature is putting both glass-box training and black-box explanation behind one API. For glass-box models the explanations are exact and editable by domain experts; for black-box models they are approximations.

It suits data scientists debugging models, auditors validating them before and after deployment, researchers comparing techniques, and business stakeholders needing transparency. Access is through pip or conda on Python, under the MIT license, with no account needed. It requires coding skills; there is no hosted app. Black-box explanations can be inaccurate, so the docs advise caution in high-stakes settings. Compute for training is your own.

What it can do

• Train glass-box models such as Explainable Boosting Machines, linear models and decision trees
• Explain black-box models using LIME, SHAP, partial dependence and Morris sensitivity
• Global and local feature importance with interactive visualizations
• Compare explanations from multiple models side by side
• Differentially private EBMs for sensitive data
• What-if analysis showing how edited inputs change predictions

Works with

• scikit-learn
• pandas
• Azure SynapseML

Example workflows

Audit a classifier for bias

Input: a tabular dataset such as adult income. Steps: split train/test, fit an ExplainableBoostingClassifier, call explain_global and show it, then inspect features like Gender or Race. Output: interactive plots showing how each feature shifts predictions.

Explain an existing black-box pipeline

Input: a trained scikit-learn pipeline (e.g. PCA plus random forest) and test data. Steps: apply a black-box explainer such as LIME or SHAP, view local explanations for selected rows. Output: approximate per-prediction feature contributions to share with stakeholders.

Before you choose

• Black-box explanations are approximate and may be inaccurate, so use caution in high-stakes settings.
• Requires Python programming; there is no hosted web app or no-code interface.
• Documentation and GitHub list different minimum Python versions, so confirm compatibility before installing.

Listed under

• Data
• Data science
• Developer tools
• Explainable AI
• Interpretability
• Machine learning
• Model debugging

Get started

Or conda install -c conda-forge interpret.

1. Install with pip install interpret or conda install -c conda-forge interpret.
2. Load your data into a pandas dataframe or numpy array and split into train and test sets.
3. Fit a glass-box model or wrap a black-box pipeline, then call explain_global or explain_local and display with show().

The package is free under the MIT license. Compute costs, including optional Azure SynapseML for distributed EBMs, are separate and not priced in the sources.

Source: github.com/interpretml/interpret, checked 28 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• InterpretML ↗interpret.ml
• GitHub - interpretml/interpret: Fit interpretable models. Explain blackbox machine learning. · GitHub ↗github.com
• Getting Started — InterpretML documentation ↗interpret.ml

Company website

interpret.ml: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.