Explainable AI-Enhanced Machine Learning Models for Early Detection of Cardiovascular Disease: Improving Predictive Performance and Clinical Transparency through Optimization

Abstract

Cardiovascular disease (CVD) is the main cause of death worldwide, highlighting an immediate requirement for enhanced diagnostic devices to trace indicators associated with CVD early in sufferers. This study leverages AI-driven machine learning (ML) algorithms to extract extrapolating more nuanced patterns from the data than traditional mortality modelling techniques. The purpose of this study is to compare and tune three ML models (Support Vector Machine (SVM), Random Forests (RF), and Logistic Regression (LR) for increasing the CVD prediction accuracy and incorporate explainable AI (XAI) techniques, such as SHAP (SHapley Additive exPlanations) for better model interpretability. The UCI Heart Disease data set is used as a backbone on which we did some very detailed level of pre-processing and then applied feature selection algorithms like PCA and RFE to select the optimal features. Optimization of models was done by hyperparameter tuning using GridSearchCV and RandomizedSearchCV. The data was 90% train test split. After split, the SVM model performed best with 94% accuracy followed by Random Forest at 92%, and Logistic Regression achieved a minimal of 90%. These findings demonstrate the value of AI-driven ML approaches can improve the prediction of cardiovascular disease. The study emphasizes that the optimized SVM model has significant potential for clinical applications in early CVD diagnosis, offering evidence that integrating AI-driven ML models into healthcare can potentially reduce global CVD mortality through earlier interventions.

Keywords: Cardiovascular Disease Prediction, Machine Learning Models, Hyperparameters Tuning, Support Vector Machine (SVM), Ensemble Learning Techniques, AI-Driven Algorithms, explainable AI (XAI), SHAP (SHapley Additive exPlanations