An Efficient Deep Learning Framework for Feature Selection in IoT-Based Heart Disease Prediction

Abstract

Heart disease remains a leading cause of morbidity and mortality worldwide, necessitating the development of efficient diagnostic tools for early detection and prevention. With the rise of large-scale medical data, leveraging advanced computational techniques such as deep learning has become critical for accurate heart disease classification. This paper proposes an efficient framework for feature extraction and feature selection techniques, utilizing deep learning models to improve classification accuracy for heart disease using Internet of Things (IoT). The framework integrates convolutional neural networks (CNNs) for feature extraction and employs a hybrid method combining Recursive Feature Elimination (RFE) with Principal Component Analysis (PCA) for feature selection. The proposed framework is evaluated on the UCI Heart Disease dataset, achieving a classification accuracy of 91.3%, precision of 90.8%, recall of 89.9%, and an F1-score of 90.3%. The model outperforms traditional methods such as SVM and random forests, which achieved accuracies of 85.2% and 86.5%, respectively. These findings demonstrate the efficacy of the proposed framework in heart disease classification and its potential for real-time clinical applications.