Research on the Impact of Green Finance on Regional Carbon Emissions and Its Mechanism under Machine Learning Framework

In this study we adopt a machine learning augmented panel model with multidimensional cluster analysis to explore the nonlinear impact mechanism of green finance on regional carbon emissions in China. A nonlinear regression model with interaction terms is constructed to capture the dynamic relationships among variables, and the XGBoost integrated learning framework is combined to enhance the model's ability to fit complex patterns, and the parameter configuration is optimized through hierarchical cross-validation to ensure the model's generalization performance. Then, the K-means clustering algorithm is used to divide the multidimensional feature space unsupervised, and the contour coefficient method is used to determine the optimal classification structure, so as to identify the clusters with significant differences in features. The SHAP value interpretation framework is finally introduced to quantify the marginal contribution of each feature to the model output, revealing the heterogeneity of the mechanism of action among variables under different clusters. This framework effectively improves the accuracy of nonlinear relationship modeling and the stability of clustering analysis while maintaining the interpretability of the model, which provides a scalable methodological paradigm for the analysis of multidisciplinary complex systems, and also provides a theoretical basis and a policy reference for the design of differentiated green financial policies.