Heart Disease Prediction Using Feature Reconstruction and Hybrid Deep Learning

Cardiovascular disorders are the leading cause of death worldwide, so it is crucial that we develop innovative, rapid and precise methods of diagnosing them. The hybrid deep learning model proposed in this study combines the most effective elements of sparse autoencoders (SAEs), multilayer perceptrons (MLPs) and convolutional neural networks (CNNs). To properly display the input attributes, this method employs categorical encoding and normalisation on the UCI Heart Disease dataset. This hybrid architecture facilitates the development of an AI system that retains crucial physiological characteristics while handling large amounts of abstract data. Furthermore, it has been demonstrated that the proposed hybrid architecture outperforms the other techniques investigated, including Learning Vector Quantification and Logistic Regression with Boruta feature selection. The hybrid design achieved a recall rate of 93.8%, an accuracy rate of 95.1%, a precision rate of 94.6% and an F1 score of 94.2%. Substantial statistical evidence supports the idea that the CNN–MLP + SAE architecture will be effective as an early detection method for heart disease.