Predicting environmental pollutants in the apartment public space: Evaluating the impact of spatial enclosure and monitoring locations

Abstract

To effectively reduce monitoring costs for pollutants in complex apartment public spaces and ensure a healthy living environment for residents, this study explores the feasibility of predicting environmental pollution in apartment public spaces through experimental and data analysis methods. By adjusting the spatial enclosure (e.g., window opening or closing) and monitoring locations on the first-floor public space, this research utilizes Spearman rank correlation coefficients and exploratory data analysis (including five linear models, nonlinear models, three tree-based models, one nearest-neighbor model, and one neural network model) to assess how these adjustments impact pollutant correlations and model performance. Results indicate that tree-based models, particularly Decision Tree Regression, consistently outperform other models, demonstrating reliable predictive capabilities across varying enclosure and location conditions. Time granularity and wind direction significantly influence correlations, while pollutants like PM and ozone exhibit unidirectional correlations across different locations. The study also finds that changes in spatial enclosure alter indoor airflow and pollutant diffusion patterns, thereby affecting predictive accuracy. Additionally, this research elucidates the feasibility of spatially predicting environmental pollutants under different conditions, offering practical guidance for low-cost monitoring of apartment public spaces. These findings support sustainable building management, effective pollution control, and enhanced health protection for residents.