Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Integrating Quantum, Deep, and Classical Features with Attention-Guided AdaBoost for Medical Risk Prediction

Rektorat UDINUS 2025 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2025
Goals
Record
271d5b72-d63e-4d31-b4ec-3883fd187c3a
Published
2026-10-30

Universitas Dian Nuswantoro conducted a medical risk prediction study with participation by Kusuma, Muh Galuh Surya Putra; Setiadi, De Rosal Ignatius Moses; Herowati, Wise; Sutojo, T.; Adi, Prajanto Wahyu; Dutta, Pushan Kumar; and Nguyen, Minh T. The activity addressed the need for accurate and interpretable diagnostic models for chronic kidney disease, diabetes, and heart disease, which are identified in the source as major causes of mortality worldwide. It examined challenges associated with conventional machine learning methods, including limited generalization, feature redundancy, and class imbalance in medical datasets.

The study proposed an integrated classification framework combining three feature paradigms: classical tabular attributes, deep latent features extracted through an unsupervised Long Short-Term Memory encoder, and quantum-inspired features derived from a five-qubit circuit implemented in PennyLane. These heterogeneous features were fused through a feature-wise attention mechanism and evaluated with an AdaBoost classifier. The attention mechanism dynamically weighted feature contributions and was used to enhance decision boundaries. Experiments covered three benchmark medical datasets involving chronic kidney disease, early-stage diabetes, and heart disease. The evaluation considered both balanced and imbalanced configurations and used stratified five-fold cross-validation. Preprocessing and feature extraction were isolated within each fold to support fair evaluation.

The proposed hybrid model outperformed conventional and ensemble baselines in the reported experiments. Peak accuracies were 99.75% for chronic kidney disease, 96.73% for diabetes, and 91.40% for heart disease, with corresponding ROC AUC values up to 1.00. Ablation analyses found that attention-based fusion improved accuracy and recall, particularly under imbalanced conditions, while SMOTE contributed minimally after feature-level optimization. The activity therefore relates to RAW_GOALS through research addressing medical risk prediction and diagnostic modelling. In institutional framing, UDINUS presents this activity as an evidence-based contribution to health research using quantum-inspired, deep, classical, attention-based, and ensemble methods.

Reference https://www.scopus.com/pages/publications/105037197025?origin=resultslist