The gradient-boosting decision tree model can predict the concentration of PAEs in children bedroom

Abstract

Exposure of phthalate has adverse effects on child health. Currently, the field measurement on PAEs concentration in children’s bedrooms were limited, and the test of PAEs is laborious. Based on the data of home detection in 454 residences from March 2013 to December 2014 in Shanghai, the association of PAEs in children's bedroom and building characteristics, residents’ lifestyle and indoor environment characterization were built by Spearman correlation. According to the Spearman correlation coefficient method, the concentration of PAEs, such as residential area was significantly correlated with DMP, BBP and DiBP in children’s bedroom (sig 0), and the use of chemicals was significantly associated with DEP and DiBP in children’s bedroom (sig 0). Then a gradient-boosting decision tree model with higher prediction accuracy is established. The influencing factors of the studied PAEs were determined by comprehensive consideration of the current study and literature review. 11 influencing factors of PAEs concentrations from three aspects were finally established in this study. The training model of GBDT has a reasonable accuracy( R2>0.9). This paper provides a reference for the prediction of PAEs concentration in the residential bedroom and the influence degree of influencing factors.