The impact of indoor air quality on work performance in urban office spaces: A machine learning approach

Abstract

Improving work performance in urban office spaces is crucial for economic growth and productivity. This study aims to explore the impact of indoor air quality on work performance and assess the relative importance of various environmental parameters by machine learning. This observational study, conducted from March to November 2022, involved a monthly questionnaire and monitoring data collection under normal office conditions without behavioral or environmental interventions. Decision tree and random forest were used to analyze the impact of environmental parameters on perceived work efficiency and work performance. The decision tree results indicated that enthalpy difference, temperature, and relative humidity emerged as the three key factors most significantly influencing perceived work efficiency, with relative importance values of 100 %, 88.2 %, and 73.8 %, respectively, in the two relatively green office spaces surveyed. This was because these factors effectively reduced the Gini index during node splitting, leading to a purer class distribution of perceived work efficiency in the dataset. In addition to direct effects, elevated levels of CO₂ and formaldehyde might further suggest the impact of pollutants on certain aspects of work performance, even in eco-friendly spaces. To optimize perceived work efficiency in summer air-conditioned offices, PM_2.5 and PM₁₀ might remain under 25 μg/m³, with CO₂ below 800 ppm. In the future, architects and operators can optimize design and operational strategies by referencing environmental parameter combinations corresponding to the top five distributions of perceived work efficiency, balancing enthalpy difference and energy consumption to improve both perceived work efficiency and energy efficiency.