Prediction of Building Carbon Emissions in Campus Areas Based on Building a Carbon Emission Correlation Factor

In the context of universalizing higher education and achieving “dual carbon goals”, accurately predicting carbon emissions from campus buildings is crucial for the development of low-carbon campuses. This study introduces a novel method for predicting carbon emissions at the regional scale for campus buildings on the basis of their functional classification. The approach begins with the manual classification of buildings within campus areas, followed by the introduction of a building carbon emission correlation factor, which accounts for the relationship between the emissions of each building type and the total emissions for that category. By applying this factor, typical building groups were identified. Three models were developed for predicting carbon emissions from campus buildings, employing mathematical statistics, machine learning, and deep learning approaches based on the building carbon correlation coefficients. The results indicated that the prediction accuracies of the SARIMA model improved by 17.23%, 6.159%, and 3.949% for buildings in categories A, B, and C, respectively, after the carbon emission correlations were considered. Similarly, the accuracies of the LSTM model improved by 2.76%, 12.636%, and 3.370%, respectively, while the accuracies of the RF model increased by 3.61%, 10.893%, and 4.776%, respectively, for these categories. These results suggest that using carbon emission correlation factors can significantly increase the overall predictive performance of these models. This study provides a reliable reference for predicting carbon emissions in campus areas, contributing to the construction of more sustainable, low-carbon campuses.