Gen-AI Assisted Machine Learning Framework for Early Heart Disease Risk Prediction

Abstract

Heart disease is a major life-threatening global health condition that often remains undetected in its early stages due to the absence of noticeable symptoms. Consequently, early risk analysis is essential for effective prevention and timely intervention. This study presents a real-world heart disease prediction system that analyzes user vitals, lifestyle and behavioral factors, and medical history to assess individual risk. The proposed system integrates machine learning with generative AI to deliver risk predictions along with personalized health insights and recommendations. The model was developed using the Heart Disease Health Indicators Dataset, comprising 253,680 records derived from over 400,000 survey responses. Five machine learning algorithms Logistic Regression, K-Nearest Neighbors, Decision Tree, Random Forest, and CatBoost were evaluated, with CatBoost achieving the highest performance, attaining a ROC-AUC score of 0.92. The application presents confidence scores from all five models to support informed decision-making. Following prediction, the results and contextual prompts are passed to a large language model (LLM) to generate actionable, personalized health recommendations. The proposed system enables real-time health assessment and decision support, particularly in emergency situations, and has the potential to facilitate early detections.

Ähnliche Arbeiten