<p>Feature-Group–Aware Cardiovascular Disease Prediction Using Gated Neural Fusion and CatBoost-Based Clinical Feature Analysis</p>

Cardiovascular (cardiac) diseases are among the leading causes of death worldwide and require effective and dependable computational methods for early diagnosis and risk assessment. The proposed study introduces a Feature-Fusion Artificial Neural Network (FF-ANN) to predict cardiovascular diseases by directly dividing the numerical and categorical clinical variables into separate learning branches and fusing them via a gated fusion process. Real-time clinical data were obtained from approximately 2,000 patient records at Ayush Multispecialty Hospital and Research Centre, Vijayapura, Karnataka, India. A leakage-free stratified validation protocol was used to evaluate the model’s performance and compare it with a CatBoost baseline and a standard ANN. The proposed FF-ANN achieved a test accuracy of 96.75%, a Matthews correlation coefficient of 0.931, and an area under the receiver operating characteristic curve of 0.996, outperforming the standard ANN. CatBoost also produced competitive results, further demonstrating the strong predictive value of categorical diagnostic variables. The findings suggest that the feature-group-aware gated fusion approach (FF-ANN) is a promising framework for analyzing heterogeneous cardiovascular clinical data to support cardiovascular decisions.