| description abstract | Effective building management strategies require a clear understanding of how occupants perceive their indoor environmental conditions. Despite their important findings, previous studies were mostly limited to single-domain evaluations of the indoor environment (e.g., thermal, visual, acoustic, or air quality), and rarely considered general well-being or productivity metrics. A holistic data analysis approach is proposed to quantify the multidomain drivers of overall comfort, perceived productivity, and perceived happiness of occupants. The approach combines exploratory and explanatory analysis methods (correlation, correspondence analysis, and machine learning) and was demonstrated using data collected from 206 occupants of 3 buildings in Abu Dhabi, United Arab Emirates. Results showed that satisfaction levels with noise, air quality, and temperature are the main drivers of the studied multidomain metrics. However, threshold-based relationships were observed at the comfort scale’s extremes, challenging the linearity assumption often adopted in previous studies. Practical implications of the findings include focusing facility management efforts on specific environmental domains that act as levers for overall satisfaction and well-being, instead of aiming to improve satisfaction with all domains simultaneously. Such levers are context-dependent, confirming the need for the proposed data analysis approach that is applicable to any built environment. Finally, the case study also highlighted the modeling capabilities of the tested machine learning algorithms (support vector machine, random forest, and gradient boosting), which achieved predictive accuracies up to 38% higher than those of regression-based statistical models. | |
