Cardiovascular disease remains a leading cause of global mortality. While Machine Learning (ML) offers powerful predictive capabilities, standard "black-box" models often lack the safety mechanisms required for healthcare deployment.
This project introduces a Hybrid Safety Architecture. It combines a high-performing Gradient Boosting classifier with a deterministic "Safety Layer" that ensures the AI never gives advice that contradicts established medical literature.
- Multi-Model Benchmarking: Comparative analysis of Logistic Regression, Random Forest, and Gradient Boosting.
- Safety Override Logic: A hard-coded medical logic layer that intercepts and corrects erroneous AI predictions.
- Interactive Smart Advisor: A tool that calculates the "minimal effective dose" of lifestyle change to lower patient risk.
- Responsible AI Audit: Verified for fairness with a negligible Equalized Odds Gap of 0.010.
| Metric | Result |
|---|---|
| Accuracy | 73.6% |
| ROC-AUC | 0.802 |
| Gender Fairness | < 0.010 |
Gender Fairness (Equalized Odds Gap) | < 0.010
To run this project on your local machine:
-
Install Dependencies: Run this command in your terminal:
pip install pandas scikit-learn matplotlib numpy -
Run the Notebook: Open
872.ipynbin Jupyter Notebook or upload it to Google Colab. -
Data Requirements: Ensure
cardio_train.csvis in the same folder as the notebook.
- 872.ipynb: Full source code and interactive tool.
- priyesh-872-finalReport.pdf: Detailed research paper.
- cardio_train.csv: Dataset (70,000 records).
Author: Priyesh Vashistha
Institution: Department of Electrical and Computer Engineering, Queen's University
Data Source: Cardiovascular Disease Dataset from Kaggle
Disclaimer: This tool is for educational purposes only and is not a substitute for professional medical advice.