Abstract
Global climate change is one of the major environmental challenges faced today, and carbon dioxide (CO(2)) emissions is the primary cause of global warming. Although many countries have committed to reducing carbon emissions and achieving carbon neutrality, progress in reducing emissions across various industries remains uncertain. To address this issue, this study employs advanced machine learning methods such as random forest, extreme gradient boosting (XGBoost), and stacked regression to construct a dynamic correlation model for exploring the relationship between sectoral electricity consumption and urban average CO(2) column concentration (XCO(2)). By combining time-rolling window techniques, the model dynamically reveals temporal correlations between sector-specific electricity consumption and urban XCO(2). The performance of the dynamic correlation model was validated with sectoral electricity consumption and urban XCO(2) data from 16 cities between 2017 and 2021. The model achieved a coefficient of determination (R(2)) of up to 0.864 and an root mean square error (RMSE) of 1.350. The study revealed that the model significantly outperformed the traditional random forest and XGBoost models in terms of prediction accuracy, effectively capturing the complex relationship between sectoral electricity consumption and carbon concentration. The correlation between electricity consumption in different industries and XCO(2) exhibited significant temporal fluctuations across cities. Through time-rolling analysis, the study revealed sector-specific influences, providing valuable insights for developing more precise and industry-targeted carbon reduction policies.