Heart Failure Prediction Using Variational Auto-Encoder and Extreme Gradient Boosted Neural Network
This paper pioneers the integration of Variational Auto-encoder (VAE) techniques into the XGBNet
model for heart failure prediction, leveraging the largest combined heart failure dataset from Kaggle.
Through rigorous evaluation over 100 epochs, the model achieved a 92% prediction accuracy in
distinguishing patients with potential heart failure. Comparative analysis revealed a 2-3% increase in
accuracy over previous methodologies, highlighting the efficacy of VAE in tandem with XGBoost. These
findings underscore the superiority of the conjugate gradient algorithm as an optimizer and its potential
implications for healthcare providers, promising enhanced accuracy in early heart failure prediction for
proactive intervention strategies and improved patient care