Author: Nikul Ram
Status: Public Open Source Project – Under Active Development
License: MIT
Repository: https://github.com/nikulram/InsightVio

InsightVio is an open-source Python toolkit for transforming machine learning model predictions into human-understandable insights. Built with a mission to improve AI transparency, trust, and decision support, InsightVio serves domains where interpretability matters most—such as healthcare, finance, and regulatory systems.

This project bridges the gap between complex AI logic and real-world practitioners by combining established explainable AI tools (like SHAP and LIME) with custom-built visualizers and interpretable logic paths.

• Visual Explanation Engine
  Converts AI decisions into intuitive visual logic trees and annotated flowcharts.

• Built-in Explainable AI (XAI) Methods
  Integrates SHAP and LIME for feature impact analysis and interpretable outputs.

• Healthcare Ready (Phase 1)
  Preconfigured to work with Breast Cancer, Heart Disease, and COVID-19 datasets.

• Modular & Research-Friendly
  Notebook-ready, lightweight, and open for research extensions.

• Production-Focused Roadmap
  API-ready and designed for clinical or regulatory embedding.

• AI Researchers and Interpretable ML Enthusiasts
• Healthcare Analysts and Medical AI Practitioners
• Ethics & Governance Professionals
• Graduate / PhD Students in AI, Bioinformatics, or Policy
• Contributors aiming for meaningful open-source impact

Install Python packages:

pip install shap lime matplotlib scikit-learn pandas numpy

Recommended: Python 3.9 or newer.(I am Using Python 3.11.5)

InsightVio currently supports classification and regression tasks on real-world healthcare datasets through both CLI and Streamlit web interface. It enables full explainability via SHAP and LIME for both uploaded CSVs and manual live input.

Currently Integrated:

• Breast Cancer Wisconsin (Diagnosis) – via load_breast_cancer() from sklearn.datasets
• Parkinson’s Disease Classification – UCI ML Repository (converted to .csv)
• Diabetes Progression (Regression) – via load_diabetes() from sklearn.datasets
• Credit Card Fraud Detection – Kaggle Dataset (converted to .csv)
• Custom CSV Upload – with automatic preprocessing and label detection
• Live Manual Input Mode – supports on-the-fly explanations using form-based entry

Upcoming (Phase 3–4):

• Cleveland Heart Disease

• COVID-19 Symptoms & Mortality Outcomes

• UCI Liver Disorders Dataset

• Model upload & selection support (external pickled models)

• Test with : streamlit run streamlit_app/app.py

All datasets are open-source and publicly available. Final README will include citations, licenses, and DOI references for proper attribution.

Interested in shaping the future of InsightVio? Collaborators are welcome for:

• Novel explainability techniques (e.g., Anchors, Integrated Gradients)
• Dataset integration (especially healthcare, finance, law)
• Academic partnerships (MS/PhD research, co-authorship, grants)

Fork this repository and open a PR, or reach out to me Nikul Ram via GitHub.

MIT License — free for personal, academic, and commercial use with attribution.

© 2025 Nikul Ram