A novel machine learning architecture to improve classification of intermediate cases in health: workflow and case study for public health
Abstract Background The practice of medicine has evolved significantly during the past decade, with the emergence of Machine Learning (ML) that offers the opportunity of personalized patient-tailored care. However, ML models still face some challenges when classifying patients where clear-cut bounda...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-07-01
|
| Series: | BMC Bioinformatics |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12859-025-06228-8 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Background The practice of medicine has evolved significantly during the past decade, with the emergence of Machine Learning (ML) that offers the opportunity of personalized patient-tailored care. However, ML models still face some challenges when classifying patients where clear-cut boundaries between classes are hard to identify. In this work, we propose an ML architecture to improve the sensitivity of detecting patients in intermediate “hard-to-classify” classes. Methods The proposed architecture replaces a single classifier with a group of cascaded increasingly specialized classifiers: the ‘Human-like’, the ‘Segregating’, and the ‘Deep’ classifiers. Its effectiveness is tested, using 8 ML algorithms (Logistic Regression, Support Vector Machine, K-Nearest Neighbor, Decision Tree, Random Forest, XGBoost, CatBoost, and Artificial Neural Network) to predict the feeling of protection among healthcare workers during the COVID-19 pandemic, based on a global online survey, then validated on two other outputs. Results The results show, for most algorithms, an enhanced detection of data points belonging to intermediate classes (up to 14% absolute increase in accuracy), as well as an overall improvement in the models’ accuracies (up to 5.8% absolute increase). The validation experiments yielded similar results with improved accuracies for most algorithms when compared to the single classifier architecture. Conclusion This novel architecture is proving to be a very promising tool for improving accuracy of the models when classifying patients in intermediate classes, regardless of the algorithm used. Accuracy-improvement for likert-type scale measures offers an opportunity for rapidly identifying “risk-profiles” during emergencies and beyond. This applies equally to patients and healthcare providers, with potential for improving quality of care and strengthening patient-centered healthcare systems that prioritize healthcare providers’ wellbeing. |
|---|---|
| ISSN: | 1471-2105 |