Study on the influencing factors on indoor PM2.5 of office buildings in beijing based on statistical and machine learning methods

To investigate the interactions and contributions of the influencing factors on indoor PM2.5, indoor monitoring data (including PM2.5, CO2, temperature, humidity, and window states) at six office buildings sites in Beijing and outdoor air quality and meteorological data at nearby national sites were collected and analyzed. Seemingly unrelated estimation (SUEST) tests and decision tree models were conducted, and results indicated that opening windows increased the infiltration factor and decreased the impact of indoor sources, while it may not be effective enough to remove indoor particles in some conditions. Besides, the influence of factors is nonlinear, and the correlations between factors make the form of cross-terms complex. The contribution of outdoor PM2.5 and periodic factors adds up to 68.43 & PLUSMN; 6.07%, and the main differences among the results are reflected in the allocation of these factors. Outdoor PM2.5 is shown to play the most important role in the variation of indoor PM2.5 in most office spaces (49.55 & PLUSMN; 16.87%). Periodic factors and personnel activities are significant (16.34 & PLUSMN; 10.02%) and strong personnel activities seemed to increase the impact of periodic factors on indoor PM2.5, while the impact of window states is limited (3.41 & PLUSMN; 2.63%). The contribution of outdoor meteorology and indoor environment is 11.89 & PLUSMN; 2.63% and 16.26 & PLUSMN; 4.40% respectively. The influence of wind direction is stronger than that of wind speed, and temperature and humidity were shown to be equally important in the variation of indoor PM2.5. The data and results can provide suggestions for indoor air quality management, and the methods proposed in this study can be used to evaluate the impact of other factors on indoor air quality based on long-term monitoring data.