Estimating the outdoor environment of workers' villages in East China using machine learning

Workers' villages in East China represent a typical form of residential government-built settlements constructed between the 1950s and the 1980s to address the housing shortage. Recent emphasis has been paid to optimizing wind and thermal comfort in older neighborhoods, following the urban renewal trend. This paper collected the geometries of 150 workers' villages. Pedestrian-level wind and Universal Thermal Climate Index (UTCI) were calculated for workers' villages using validated simulation software. Seven machine learning (ML) algorithms were compared for modeling the nonlinear relationship between the building morphology and the outdoor environment of the workers' villages. The ensemble model, especially the Adaboost model, performs best when predicting static wind ratio and UTCI with R2 values of 0.89 and 0.99. The trained models were applied to es-timate the outdoor environment of 1118 workers' villages in East China. The result shows most workers' villages have static wind ratios over 0.7. Workers' villages in Jiangsu endure more extreme summer heat, whereas workers' villages in Zhejiang have a higher static wind ratio in winter and summer. The use of ML offers a quicker estimation of outdoor wind and thermal comfort in large-scale workers' villages than numerical simu-lations, therefore shedding light on the targeting of urban renewal.

Automated summary

This study collected geometries of workers' villages and used machine learning algorithms to model and estimate outdoor wind and thermal comfort. The results showed that most workers' villages have high wind ratios. Additionally, workers' villages in Jiangsu experience more extreme summer heat, while those in Zhejiang have higher wind ratios in winter and summer.