Novel Deep Neural Network for Early Prediction and Prevention of Cardiovascular Disease

Abstract

Abstract Cardiovascular diseases (CVD) are common and fatal conditions requiring early detection for reduced mortality rates. Machine learning algorithms hold promise for identifying risk factors. This study presents a comprehensive system for efficient CVD prediction and prevention. Accurate training data is generated through real-time datasets, preprocessing, and hybrid dataset creation (Cleveland, VA Long Beach, Switzerland, Hungarian, and Stat log). Feature selection optimizes prediction, including ANOVA and CHI2SQUARE methods. Classifier models (Decision Tree, Random Forest, KNN, Naïve Bayes, SVM, DNN) are trained on the hybrid dataset using class balancing and feature selection. DNN with CHI2-Square selection achieves 99.27% accuracy; CBFS-DNN on real-time data reaches 82.06%. The ongoing research develops a prevention model focusing on ten key features, aiding early CVD risk identification and tailored interventions. The system's rapid prediction in 0.05 seconds enables timely preventive actions.