Application of Machine Learning Algorithms for Geogenic Radon Potential Mapping in Danyang-Gun, South Korea

Continuous generation of radon gas by soil and rocks rich in components the uranium chain, along with prolonged inhalation progeny enclosed spaces, can lead to severe respiratory diseases. Detection radon-prone areas acquisition detailed knowledge regarding relationships between indoor variations geogenic factors facilitate implementation more appropriate mitigation strategies high-risk residential zones. In present study, 10 (i.e., lithology; fault density; mean calcium oxide [CaO], copper [Cu], [Pb], ferric [Fe 2 O 3 ] concentrations; elevation; slope; valley depth; topographic wetness index [TWI]) were selected map potential based on measurements levels 1,452 dwellings. Mapping was performed using three machine learning methods: long short-term memory (LSTM), extreme (ELM), random vector functional link (RVFL). The results validated terms area under receiver operating characteristic curve (AUROC), root square error (RMSE), standard deviation (StD). prediction abilities all models satisfactory; however, ELM model had best performance, AUROC, RMSE, StD values 0.824, 0.209, 0.207, respectively. Moreover, approximately 40% study covered very high zones that mainly included populated Danyang-gun, South Korea. Therefore, be used establish construction regulations radon-priority areas, identify cost-effective remedial actions for existing buildings, thus reducing and, extension, exposure-associated effects human health.