Nowadays, climate change has become a serious issue for the world, and the creation of low-carbon green buildings becomes one of the ways offered by humans to overcome the existing problems. In this research, through BIM technology, inventory data and emission data from selected buildings have been collected. With correlation analysis and elastic net algorithm, design features affecting building carbon emissions have been screened out, and 8 features were regarded as predictors. Then, an improved gray wolf optimization algorithm-based support vector machine method (IGWO-SVM) is utilized to establish the prediction model of building carbon emissions. Through model comparisons, it has been found that our IGWO-SVM model has attained an R² value of 0.811, which is 9.45% to 125.91% better than other models, while the metrics of RMSE, MAE, NRMSE, and CV(RMSE) have reached the lowest values compared to other models with at least 12.26% lower performance. It will help architects estimate carbon emissions accurately, enabling green and low-carbon buildings to be promoted.
