Artificial Intelligence and Machine Learning-Based Prediction of Cardiovascular Disease Risk in Rural Elderly Women in North-East India: Insights into Women's Health, HDL, Metabolic Syndrome, and Key Biomarkers

Background: This study aimed to leverage artificial intelligence and machine learning approaches to predict cardiovascular disease (CVD) risk among rural postmenopausal women in North-East India. The objectives included identifying critical risk factors and evaluating the performance of various machine learning models to enhance CVD prediction accuracy in rural populations. Methods: The research was conducted in Singur Block (West Bengal) and Amdanga Block (North 24 Parganas District) over a span of four years, from March 2014 to August 2018. Key risk factors, including waist circumference, blood pressure, fasting blood glucose, and HDL cholesterol, were collected and analyzed. A correlation matrix was employed to assess the associations between these variables and CVD risk. Additionally, eight machine learning models were deployed to predict CVD risk, with performance comparisons to identify the most effective algorithm. Results: The correlation analysis revealed strong associations between CVD risk and factors such as waist circumference, blood pressure, and fasting blood glucose, with HDL cholesterol playing a lesser yet relevant role. Among the machine learning models tested, the Random Forest algorithm achieved the highest accuracy of 98.91%, followed by Decision Tree and Support Vector Classifier (SVC), both of which achieved accuracies exceeding 95%. Feature important analysis highlighted waist circumference as the most significant predictor, with blood pressure and fasting blood glucose also contributing substantially to the prediction of CVD risk, while HDL cholesterol had a comparatively lower but still notable impact. Conclusion: Machine learning techniques, particularly the Random Forest model, demonstrated remarkable efficacy in predicting CVD risk among rural postmenopausal women. This study identifies waist circumference, blood pressure, and fasting blood glucose as key predictors, emphasizing the potential of machine learning-based approaches as cost-effective, scalable solutions for early CVD detection. These findings provide a foundation for targeted preventive measures and improved health outcomes in underserved rural settings.