Multi-Faceted Approach to Cardiovascular Risk Assessment by Utilizing Predictive Machine Learning and Clinical Data in a Unified Web Platform

Cardiovascular diseases (CVD) persist as a formidable global health challenge, underscoring the imperative for advanced early detection mechanisms. The evolution of computational methods within healthcare has paved the way for transformative applications of machine learning, offering solutions that enhance diagnostic accuracy and contribute to the SDG-3; Good Health and Well-Being. This study aims to identify an algorithm with consistent performance across diverse datasets and integrate it into a comprehensive and user-centric approach to heart disease prediction. The investigation includes an examination of eight machine learning algorithms, three deep learning algorithms, and four heterogeneous datasets from the Kaggle. These algorithms' predictive performance is assessed through Precision, Recall, F1 score, Accuracy, and Area Under the Curve (AUC). A Principal Component Analysis (PCA) feature engineering approach is presented to boost predictive performance. An alternative feature selection method, Lasso, was explored, with PCA emerging as the optimal choice for accuracy in the given datasets. As such, the XGBoost algorithm with PCA achieves an impressive accuracy rate and F1 score of around 99% along with an excellent 97% AUC rate in disease prediction on the other dataset. The selected XGBoost model is integrated into a user-friendly web application, providing a holistic platform for heart disease management. Furthermore, we recommended an RPA, IoMT, and AI-based tailored solution to make our web application more reliable, which we have proven in our study is attainable.