DEVELOPING A DATA-DRIVEN THERMAL AND ACOUSTIC COMFORT PREDICTION MODEL FOR A MIXED-MODE BUILDING IN SINGAPORE

Abstract

This study aims to develop a data-driven thermal and acoustic comfort prediction model for a mixed-mode building in Singapore. The research question is whether a Random Forest Classifier can predict thermal and acoustic comfort as accurately as the Predicted Mean Vote model. To collect data, an experiment was conducted under mixed-mode ventilation in an education institution and surveys were conducted in office spaces to gather information about participants' thermal sensation, thermal preference, thermal acceptability, and acoustic sensation. Selected data was then input into the random forest classifier to predict thermal and acoustic comfort. The proposed model was able to predict Thermal Sensation Vote within -3 to +3 with similar accuracy to the PMV model. However, the Random Forest Model outperformed the PMV model in predicting TSV within the range of -3 to -1. Additionally, the proposed model had approximately a 50% accuracy in predicting Acoustic Sensation Vote despite the presence of an imbalanced dataset. The study concludes that it is possible to predict thermal and acoustic comfort with fewer inputs than required by the PMV model, with potential applications in the field of mixed-mode ventilation.

Overall, this study demonstrates the potential of using a Random Forest Classifier to predict thermal and acoustic comfort in mixed-mode buildings in Singapore. By comparing the proposed model to the PMV model, this research highlights the advantages and limitations of different approaches to predicting thermal and acoustic comfort. The findings suggest that the proposed model is a viable alternative to the PMV model, particularly for predicting TSV within the range of -3 to -1. Future research could explore ways to improve the accuracy of the proposed model, as well as its applications in other settings beyond office spaces.