An effective deep learning-based ensemble model for heart disease prediction

doi:10.1007/s00500-025-10907-2

An effective deep learning-based ensemble model for heart disease prediction

2025-09-23

0

PRE

AI

DOI：10.1007/s00500-025-10907-2

Original

Original request for help

Save

Abstract

En 中文

This research proposes a Deep Learning (DL) based ensemble model combined with Synthetic Minority Oversampling Technique (SMOTE) and correlation Features Selection Technique (FST) for HD prediction. The Cleveland dataset was subjected to pre-processing steps and was balanced using SMOTE, which also minimizes the noise. The Pearsons Correlation Coefficient (PCC) was applied to select the relevant and most informative features based on a strong correlation level of 0.2 between input attributes and target classes. The DL Stacked Ensemble Learning (EL) classifier (DL-SEL) was formed by combining classifiers, which were optimized for obtaining best performances and was evaluated using metrics for comparative analysis. The stacked DL ensemble method with SMOTE and PCC obtained the optimal performance in comparison with the various Machine Learning (ML) classifiers and previous standard works. The stacked ensemble of DNN and RF with SVC as meta-learner obtained 95.38% and 90.29% accuracy in multiclass and bi-class modes respectively.

Keywords:

Cardio vascular diseases

SMOTE

Features Selection

Deep Learning

Ensemble Learning

An effective deep learning-based ensemble model for heart disease prediction

Abstract

Journal

Organization