.png){kind=icon}
.png){kind=icon}
Return
Data-driven approach to develop prediction model for outdoor thermal comfort using optimized tree-type algorithms
.png){kind=icon}
PRE
AI
DOI：10.1016/j.buildenv.2022.109663.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Thermal comfort can affect the productivity, health, and satisfaction of people. Although indoor thermal comfort can be controlled using heating, ventilation, and air conditioning, this is difficult for outdoor thermal comfort. Therefore, it is important for evaluating outdoor thermal comfort to manage the health and productivity of people for a specific industry, such as construction. However, conventional simulations are very difficult to conduct by non-experts. Moreover, in previous studies, simplified models have low prediction accuracy. To solve these issues, this study develops a user-friendly data-driven prediction model that maximizes prediction accuracy using an optimized tree-based machine learning algorithm. This data-driven prediction model construction for outdoor thermal comfort using machine learning is made up of three steps: (i) establishment of a database, (ii) selection of variables, and (iii) selection of prediction model. This study considers three scenarios to maximize the prediction accuracy. The results reveal that the highest prediction accuracy (95.21%) is achieved using the XGBoost algorithm. Moreover, five-fold cross-validation is conducted to validate the prediction model. It shows that the developed prediction model can accurately predict outdoor thermal comfort. Additionally, non-experts can collect input data from a public institution or a sensor and easily utilize the prediction model.
Keywords:
Physiological equivalent temperature
Prediction model
XGBoost
Hyperparameter tunning
Journal
IF:
7.6
Papers:
1.3W
Citations:
6.6W
