An Ensembled Hybrid Machine and Deep Learning Approach for Heart Diseases Prediction

The early detection and treatment of heart disease is crucial for reducing mortality rates from this ailment, which is a leading cause of death worldwide. If doctors could predict which patients would most likely develop heart disease in advance, they might tailor their preventative measures and treatments to those individuals. There are various methods for heart disease, but assessment accuracy remains one of the big challenges in this field. This study aims to introduce a predictive model for heart disease diagnosis. In this paper, we developed an ensemble hybrid machine and deep learning approach based on a two-stage experiment on the integrated heart disease dataset of five heart disease datasets from the UCI-ML warehouse. First, using the recursive feature elimination technique, we tested thirteen machine learning and deep learning algorithms on the featured dataset. In the second stage, we applied basic ensemble learning approaches, including weighted averaging, hard voting, and soft voting, to the first approaches and observed the enhanced predictive outcomes with the same dataset with the same integrated dataset as 98.32% accuracy, 99.43% precision, 98.31% sensitivity, 98.33% specificity, and 98.33% F1-score that justifies the novelty of the proposed model. Finally, ROC curves with AUC values were added to the study's empirical analysis, and an improved AUC of 0.979 was recorded. This proposed work can help physicians diagnose heart disease.