Adaptive confidence-aware ensemble machine learning for viral disease detection and severity

William Mary, Albert Antony Raj

Abstract


The rapid emergence and overlapping clinical presentations of viral infectious diseases such as COVID-19 and dengue pose significant challenges for timely diagnosis and risk stratification in resource-limited settings. Conventional single-model approaches often suffer from limited generalizability, class imbalance, and suboptimal feature utilization, resulting in delayed interventions and inefficient resource allocation. This study introduces a weighted adaptive confidence-aware ensemble machine learning framework (WACAM) for accurate early detection and subsequent severity prediction of viral infections using routine hematological and demographic data. A merged dataset comprising 1,369 patient records was preprocessed with median imputation, Z-score outlier handling, Winsorization, and standard scaler normalization. Boruta feature selection identified key clinical markers, followed by synthetic minority over-sampling technique (SMOTE) for class balancing. Seven diverse base learners were integrated through a novel confidence-weighted soft-voting scheme. Model weights were adaptively assigned via exponential softmax of out-of-fold (OOF) area under the curve (AUC) scores, with an optimized probability threshold maximizing Matthews correlation coefficient (MCC). The proposed WACAM achieved superior performance with 86.5% accuracy, 0.8795 F1-score, 0.9156 ROC-AUC, and 0.726 MCC on the test set, outperforming individual models.

Keywords


Confidence-aware soft voting; Multi-disease classification; Severity prediction; Viral disease detection; Weighted adaptive ensemble

Full Text:

PDF


DOI: https://doi.org/10.11591/eei.v15i6.13965

Refbacks

  • There are currently no refbacks.


Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Bulletin of EEI Statistics

Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191 , e-ISSN: 2302-9285
This journal is published by the Institute of Advanced Engineering and Science (IAES) .