A Hybrid CNN-BiLSTM Model for Heart Disease Prediction on Cloud-Hosted Health Data

Abstract

Heart disease remains a leading cause of mortality worldwide, necessitating advanced predictive models for early detection and prevention. Traditional machine learning techniques, though widely used, often struggle with feature dependencies and sequential patient data analysis. This study proposes a Hybrid Convolutional Neural Network - Bidirectional Long Short-Term Memory (CNN-BiLSTM) model for heart disease prediction. The CNN component extracts high-level spatial features from clinical parameters, while the BiLSTM captures temporal dependencies in patient records, enhancing classification performance. The model is trained and evaluated on the Heart Disease Ensemble Classifier dataset, achieving an accuracy of 99.57%, precision of 99.48%, recall of 99.65%, and an F1-score of 99.56%. Furthermore, the AUC-ROC score of 0.9968 and average precision (AP) of 0.9970 demonstrate superior classification capability. Comparative analysis against traditional classifiers such as Support Vector Machines (SVM), Decision Trees, and Random Forest highlights the effectiveness of the proposed approach. The results indicate that the hybrid deep learning model significantly improves heart disease prediction by reducing false negatives and increasing overall reliability.