IMPROVING CARDIOVASCULAR DISEASE PREDICTION WITH DEEP LEARNING AND CORRELATION AWARE SMOTE

Abstract

Cardiovascular disease (CVD) remains a leading cause of mortality worldwide, necessitating accurate and efficient predictive systems for early diagnosis. This study proposes a hybrid ensemble framework that integrates machine learning and deep learning models to enhance prediction performance. The methodology utilizes the cardio_train dataset, consisting of clinical and lifestyle attributes, and applies preprocessing techniques such as feature engineering, standardization, and correlation-aware synthetic oversampling to handle cla ss imbalance. The model combines XGBoost, Random Forest, and an Artificial Neural Network (ANN) within a stacking/ensemble approach, where predictions are aggregated using probability averaging. Feature selection is incorporated to reduce noise and improve computational efficiency. The system is trained using optimized hyperparameters, with ANN employing ReLU activation, Adam optimizer, and binary cross-entropy loss. Experimental results demonstrate improved accuracy, robustness, and generalization compared to individual models, with reliable classification of cardiovascular risk levels (low, moderate, high). Impressive accuracy of 99.08% and F1-score of 99.53%, demonstrating the effectiveness of combining spatial and temporal learning. The proposed approach enhances prediction consistency and supports clinical decision-making through confidence-based outputs, making it a practical tool for real-world healthcare applications.