Determinants and Prediction of Classroom Air Quality: A Logistic Regression Approach

Abstract

Classroom air quality is an important determinant of students' health, comfort, and academic performance. This study investigated the factors influencing classroom air quality and evaluated the predictive ability of selected environmental and occupancy-related variables in distinguishing between moderate and non-moderate air quality conditions. A dataset comprising 500 valid observations was analyzed using descriptive statistics, multicollinearity diagnostics, binary logistic regression, and Receiver Operating Characteristic (ROC) curve analysis. The predictor variables included estimated student count, carbon dioxide (CO₂) concentration, particulate matter (PM₂.₅) concentration, temperature, relative humidity, robot X position, robot Y position, school period, and ventilation decision. Descriptive results revealed considerable variability in classroom occupancy and pollutant concentrations, while temperature and humidity remained relatively stable. Variance Inflation Factor (VIF) analysis indicated severe multicollinearity among CO₂ concentration (VIF = 29.331), PM₂.₅ concentration (VIF = 19.262), humidity (VIF = 18.187), and ventilation decision (VIF = 11.759). The logistic regression model demonstrated excellent predictive performance, achieving 100% classification accuracy and a Nagelkerke R² value of 1.000. However, none of the predictor variables were statistically significant (p > 0.05), and the estimation process failed to converge, suggesting possible complete or quasi-complete separation and instability in the parameter estimates. ROC analysis identified CO₂ concentration (AUC = 0.999), PM₂.₅ concentration (AUC = 0.995), relative humidity (AUC = 0.994), estimated student count (AUC = 0.942), and temperature (AUC = 0.812) as strong predictors of classroom air quality status. In contrast, robot X position (AUC = 0.405) and robot Y position (AUC = 0.529) exhibited weak discriminatory ability. The findings highlight the importance of environmental pollutants and occupancy-related factors in predicting classroom air quality conditions.