Machine learning prediction model for medical environment comfort based on SHAP and LIME interpretability analysis
Changsheng Zhang1, Linjun Liu2
1Faculty of Artificial Intelligence and Big Data, ZiBo Polytechnic University, Zibo, China.
Scientific Reports
|November 10, 2025
Summary
Machine learning accurately predicts patient discomfort using environmental data. Air quality and temperature are key factors influencing comfort in medical settings, guiding personalized environmental control strategies.
Area of Science:
- Environmental Science
- Medical Informatics
- Machine Learning
Background:
- Patient comfort is crucial for treatment outcomes and recovery.
- Medical environments require optimization for patient well-being.
- Environmental factors significantly impact patient experience.
Purpose of the Study:
- To develop a machine learning model for predicting patient discomfort.
- To identify key environmental factors affecting patient comfort in medical infusion rooms.
- To provide a scientific basis for intelligent medical environment management.
Main Methods:
- Collected 1,000 samples with 11 environmental features.
- Compared 10 machine learning algorithms, selecting XGBoost.
- Utilized SHAP and LIME for model interpretability.
Main Results:
- XGBoost model achieved 85.2% accuracy.
- Air quality index and temperature were identified as most critical factors.
- Humidity and noise level also significantly impacted patient discomfort.
Conclusions:
- Machine learning effectively predicts patient discomfort from environmental data.
- Interpretability analysis revealed specific influence mechanisms of environmental factors.
- Findings support personalized environmental control and intelligent medical facility management.
