Tianqi Yu, Xiaoqi Zhan, Zichu Tian, Daoru Wang

A Machine Learning-Based Approach to Evaluate the Spatial Performance of Courtyards—A Case Study of Beijing’s Old Town

  • Building and Construction
  • Civil and Structural Engineering
  • Architecture

The quality of residential buildings in old urban areas of Beijing is known to be inconsistent, prompting numerous urban renewal projects in the city. This research investigates how building space impacts energy usage and daylighting in courtyard areas of old urban regions in northern China. It also proposes a quick evaluation method for building performance in courtyard spaces, utilizing multi-objective optimization and machine learning classification prediction as a theoretical framework. A study was conducted to gather and organize building space parameters and their corresponding performances using a genetic algorithm. The dataset was then pre-processed and trained using the LightGBM algorithm. The model validation results revealed a recall of 0.9 and an F1-score of 0.8. These scores indicate that the design scheme’s performance level can be accurately identified in practical use. The goal of this study is to propose a set of rapid assessment methods for building performance levels in courtyard spaces. These methods can significantly improve the feedback efficiency between design decision and performance assessment, reduce the time wasted in building performance simulation during the architectural design process, and avoid unreasonable renovation and addition in urban renewal. Furthermore, the research method has universality and can be applied to courtyard-shaped buildings in other regions.

Need a simple solution for managing your BibTeX entries? Explore CiteDrive!

  • Web-based, modern reference management
  • Collaborate and share with fellow researchers
  • Integration with Overleaf
  • Comprehensive BibTeX/BibLaTeX support
  • Save articles and websites directly from your browser
  • Search for new articles from a database of tens of millions of references
Try out CiteDrive

More from our Archive