ExplainX is an open-source Python framework designed to provide explainable AI (XAI) capabilities for data scientists and business users. It aims to simplify the process of understanding, debugging, and validating machine learning models by offering a unified interface for various interpretability techniques, enabling better trust and deployment of AI solutions.

How It Works

ExplainX integrates multiple XAI techniques, including SHAP (Kernel and Tree Explainer), Partial Dependence Plots, and What-If Scenario Analysis, into a single framework. Users can apply these techniques with a single line of code after model training. The framework automatically detects model and problem types (classification/regression), streamlining the analysis process and generating interactive dashboards for easy sharing and comprehension of model behavior.

Quick Start & Requirements

• Install via pip: pip install explainx
• Requires Python 3.5+. For Windows users, Microsoft C++ Build Tools must be installed first.
• Supports cloud environments like AWS SageMaker and Colab.
• Official Demo: http://3.128.188.55:8080/
• Cloud installation guide available.

Highlighted Details

• Supports a wide range of models including Scikit-learn, XGBoost, Catboost, SVM, and H2O.ai AutoML.
• Offers both a complete dashboard view and individual module access for specific analyses.
• Includes features like model metrics, global SHAP values, and cohort analysis for model performance comparison.
• "What-If" scenario analysis allows for local-level explanation of predictions.

Maintenance & Community

The project is actively seeking co-authors to further development. Contact is available via ms8909@nyu.edu. Contribution guidelines are provided, encouraging forking and pull requests, with a process for discussing major changes.

Licensing & Compatibility

Licensed under the MIT License, permitting commercial use and integration with closed-source projects.

Limitations & Caveats

TensorFlow and PyTorch model support are listed as "Coming Soon." Surrogate Decision Trees and Anchors are also planned future additions. The project is seeking contributors, which may indicate a smaller core development team.