A performance-driven weighted soft-voting ensemble for enhanced heart disease prediction: validation on the UCI benchmark
Journal of Statistical Sciences and Computational Intelligence · 30.06.2026 · 2 (2) · 691-704
DOI: 10.64497/jssci.232 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Statistical Sciences and Computational Intelligence, online erschienen am 30.06.2026. Verfasst von Aina'u Nadada und 2 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Cardiovascular diseases (CVDs) represent the leading cause of global mortality, with early detection crucial for reducing preventable deaths. While machine learning (ML) shows promise in cardiac risk prediction, existing approaches often rely on single classifiers that are susceptible to model-specific biases and exhibit inconsistent false-negative rates in clinical settings. This study proposes a performance-driven weighted soft-voting ensemble framework that intelligently integrates five ML classifiers: Random Forest, Bagged Trees, XGBoost, Support Vector Machine, and Artificial Neural Network. Utilizing the combined UCI Heart Disease dataset (n=569), classifier weights were assigned proportionally to validation ROC-AUC scores through a systematic optimization process. The architecture employs a dual-layer strategy encompassing model-level hyperparameter tuning and ensemble-level composition selection. Through stratified 10-fold cross-validation, the optimized ensemble achieved competitive performance metrics: accuracy = 0.9649, precision = 0.9579, recall = 0.9653, F1-score = 0.9615, and ROC-AUC = 0.9912. The model demonstrated high clinical safety with only 9 false negatives and 11 false positives. Ablation analysis confirmed that when base classifiers exhibit similar validation performance, weight granularity yields marginal gains; instead, the framework's systematic performance-based model selection provides transparent, reproducible criteria for clinical ensemble construction. This approach offers a robust, clinically aligned alternative to conventional equal-weight ensembles, effectively balancing predictive accuracy with diagnostic safety for resource-constrained healthcare environments.
Zitieren
APA
Nadada, A., Mu'azu, A. A., & Zubairu, B. (2026). A performance-driven weighted soft-voting ensemble for enhanced heart disease prediction: validation on the UCI benchmark. Journal of Statistical Sciences and Computational Intelligence, 2(2), 691-704. https://doi.org/10.64497/jssci.232
Vancouver
Nadada A, Mu'azu AA, Zubairu B. A performance-driven weighted soft-voting ensemble for enhanced heart disease prediction: validation on the UCI benchmark. Journal of Statistical Sciences and Computational Intelligence. 2026;2(2):691-704. doi:10.64497/jssci.232
BibTeX
@article{Nadada2026,
author = {Nadada, Aina'u and Mu'azu, Abubakar Aminu and Zubairu, Babangida},
title = {A performance-driven weighted soft-voting ensemble for enhanced heart disease prediction: validation on the UCI benchmark},
journal = {Journal of Statistical Sciences and Computational Intelligence},
year = {2026},
volume = {2},
number = {2},
pages = {691--704},
doi = {10.64497/jssci.232}
}
EUVIMED