Machine learning models for predicting indoor airborne fungal concentrations in public facilities utilizing environmental variables

Airborne fungi are major contributors to substandard indoor air quality, with potential implications for public health, especially in public facilities. The risk of chronic exposure can be significantly reduced by accurately predicting airborne fungal concentrations. To manage indoor air quality, we developed machine learning (ML) models that predict airborne fungal concentrations in public facilities by utilizing environmental variables, such as facility type, floor, month, air temperature, relative humidity, coarse particulate matter (PM)_2.5–10, and 2-day accumulated precipitation. A gene-based assay with high specificity and sensitivity was used to measure the fungal concentrations. The Gradient Boosting (GB) model exhibited superior performance among the seven developed models, achieving an R² of 0.78 on the test set. SHapley Additive exPlanations (SHAP) analysis was performed to evaluate the significance of the features. According to our findings, day care centers had the most substantial influence on fungal concentrations compared to those of other facilities. The impact of the 2-day accumulated below-average precipitation was more significant than that of extreme precipitation in increasing fungal concentrations. Furthermore, fungal concentrations were positively correlated with air temperature, coarse PM_2.5–10, and relative humidity. Based on these findings, we may provide fundamental insights into airborne fungal concentrations and the environmental variables that influence them, while the GB model developed herein can serve as a tool for assessing microbial contamination in public facilities.