Related Experiment Video
Updated: Sep 11, 2025

Measuring Carbon-based Contaminant Mineralization Using Combined CO2 Flux and Radiocarbon Analyses
Published on: October 21, 2016
Decoding indoor radon: An explainable AI approach to quantifying building, environmental, and inhabitants'
G Gigante1, S Antignani1, C Di Carlo1
1Italian National Institute of Health, National Center for Radiation Protection and Computational Physics, viale Regina Elena 299, I-00161 Roma, Italy.
Explainable AI significantly improved indoor radon concentration prediction, identifying building and inhabitant factors. Further research is needed to understand unexplained radon variability.
Area of Science:
- Environmental Health
- Artificial Intelligence
- Geospatial Analysis
Background:
- Indoor radon exposure is a leading cause of lung cancer, with significant unexplained variability in indoor radon concentration (IRC).
- Traditional models fail to capture the multifactorial origins of IRC, necessitating advanced analytical approaches.
Purpose of the Study:
- To enhance the prediction of indoor radon concentration (IRC) using explainable AI (XAI).
- To quantify the contributions of building, environmental, and inhabitant characteristics to IRC variability.
- To provide guidance for targeted radon risk mitigation strategies.
Main Methods:
- Utilized the XGBoost algorithm for IRC prediction on a high-quality dataset from the Second Italian National Radon Survey.
- Applied SHAP (SHapley Additive exPlanations) algorithm to analyze feature contributions and understand model predictions.
- Performed sensitivity analysis and compared results with a linear model (Lin-NN) for robustness.
Main Results:
- Achieved significantly improved IRC prediction performance compared to previous models.
- Quantified the relative importance of building, environmental, and inhabitant characteristics in explaining IRC.
- Identified a notable residual variance, likely due to unobserved factors like inhabitant behavior.
Conclusions:
- XAI, particularly XGBoost and SHAP, offers a powerful approach to understanding and predicting indoor radon concentration.
- Inhabitant characteristics play a crucial, often overlooked, role in IRC variability.
- Findings support the development of targeted radon mitigation strategies and potential building code updates.
Related Concept Videos
Biological Effects of Radiation
Measurement of Air Content in Concrete
The pressure method,...
Quantifying Heat
Radioactive Decay and Radiometric Dating
Real Time RT-PCR
The real-time quantification of the number of amplified products is...
Radiation Pressure: Problem Solving
The average value of the rate of momentum transfer divided by the absorbing area represents the average force...

