An Ensemble Deep Learning Model to Enhance Heart Disease Prediction

Abstract

In medicine, heart disease prediction is important since it can lower health risks. Deep learning (DL) and traditional machine learning (ML) methods have been used in numerous research to predict heart disease. Current methods need to be enhanced and improved in order to increase the prediction accuracy and performance for heart disease. An optimized stacking ensemble model was suggested in this research to enhance the accuracy of heart disease prediction. The proposed model integrates three DL models: Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) that were used in combination with two meta learners: Support Vector Machine and Logistic Regression (LR) to enhance the performance of the model. The Cleveland dataset was applied with two feature selection methods: chi-squared and Recursive Feature Elimination (RFE). The proposed model was compared with single DL and the four standard ML methods: SVM, LR, Decision Tree (DT), and Naive Bayes (NB). The DL and ML parameters were iteratively optimized using Keras-tuner and grid search. In order to estimate the effectiveness of the models and validate the findings, accuracy, recall precision, and f1-score were used. The results show that the proposed model with full features dataset has performed the highest performance compared with single DL and ML models. The suggested model provided better results than the existing heart disease prediction models.