Ensemble Learning Approaches for Automatic Prediction of COPD Based on Clinical Data
Digital Health Research · 2024
DOI: 10.61499/dhr.2024.2.e4 ↗Kurz gefasst
Fachartikel in der Zeitschrift Digital Health Research, 2024. Verfasst von Tae Gwan Jang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 20 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: Ministry of Health and Welfare, National Research Foundation of Korea, Digital Healthcare Industry in Gangwon province.
Themen: COPD
Abstract (Originalsprache)
Background
Chronic obstructive pulmonary disease (COPD) is a chronic respiratory condition mainly caused by environmental factors like smoking and air pollution, leading to persistent symptoms and worsening lung function.Early detection is vital to prevent irreversible damage, but current methods struggle to identify COPD in its early stages.
Methods
In this study, we develop ensemble learning models to automatically predict COPD by leveraging clinical data.Ensemble learning approaches for COPD detection were utilized with four main algorithms: random forest (RF), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), and categorical boosting (CatBoost).Each algorithm was trained and validated using clinical datasets extracted from a total of 886 subjects (control: 443, COPD: 443).The clinical data included approximately 1,500 variables, covering pulmonary function tests, demographic information, and medical history.The model's output was a binary classification task to distinguish between healthy individuals and COPD patients.Additionally, the study identified the most important features contributing to COPD prediction, ranking them based on their influence in each ensemble learning model.
Results
Our results showed that the XGBoost algorithm achieved the best performance for COPD detection, with an accuracy of 78% and an area under the curve of 0.86.In addition, a SHapley Additive exPlanations summary plot was used to visualize the impact of features across all models.
Conclusion
The proposed ensemble learning models, particularly XGBoost and LightGBM, are effective for automatically predicting COPD.These models outperform traditional predictive models in handling complex data patterns and interactions.Further studies should validate these results in diverse populations and integrate additional data, such as genetic or environmental factors, to enhance predictive performance.
Förderung
- Ministry of Health and Welfare
- National Research Foundation of Korea
- Digital Healthcare Industry in Gangwon province
Zitieren
APA
Jang, T. G., Park, S. Y., Park, H. Y., Lee, J. A., Kim, S., & Urtnasan, E. (2024). Ensemble Learning Approaches for Automatic Prediction of COPD Based on Clinical Data. Digital Health Research. https://doi.org/10.61499/dhr.2024.2.e4
Vancouver
Jang TG, Park SY, Park HY, Lee JA, Kim S, Urtnasan E. Ensemble Learning Approaches for Automatic Prediction of COPD Based on Clinical Data. Digital Health Research. 2024. doi:10.61499/dhr.2024.2.e4
BibTeX
@article{Jang2024,
author = {Jang, Tae Gwan and Park, So Yeon and Park, Ha Young and Lee, Ji Ae and Kim, Sang-Ha and Urtnasan, Erdenebayar},
title = {Ensemble Learning Approaches for Automatic Prediction of COPD Based on Clinical Data},
journal = {Digital Health Research},
year = {2024},
doi = {10.61499/dhr.2024.2.e4}
}
EUVIMED