Optimization and prediction in the early design stage of office buildings using genetic and XGBoost algorithms

Incorporating intelligent optimization algorithms in the early stages of office building design facilitates a better response to the local climate. The indoor and outdoor thermal performances of office buildings, such as solar radiation, indoor lighting, and outdoor thermal comfort, must be jointly evaluated during the conceptual design phase. Based on the technical framework of performance-based generative architectural design, this study constructs a data-driven workflow for comprehensive performance assessment and rapid prediction of office buildings. The method was then applied to an office building in the hot summer and cold winter regions of China. Based on a total of 6000 data samples generated by the iterative process of genetic optimization, this study achieved a precision of 0.77, recall of 0.59, and F-1 score of 0.75 for categorical prediction by the XGBoost algorithm. The method facilitates the optimization potential of integrated solar and thermal performances in the early design phase of office buildings while significantly improving the efficiency of interaction and feedback between design decisions and their performance evaluation.

Automated summary

Incorporating intelligent optimization algorithms in the early stages of office building design allows for better adaptation to local climates and improved indoor and outdoor thermal performances. This study utilizes a data-driven workflow based on performance-based generative architectural design to comprehensively assess and rapidly predict the performance of office buildings. By generating 6000 data samples through an iterative process of genetic optimization, this study achieved high precision and recall in categorical prediction using the XGBoost algorithm.