ESG Digital Evaluation System for Small and Medium-sized Enterprises

In response to the technical pain points of unstructured ESG information, fragmented data, uneven disclosure quality, insufficient robustness of traditional rating methods, and difficulty in adapting to industry differences, this paper constructs an end-to-end ESG digital evaluation and green finance decision support system. The system takes multi-source heterogeneous data as input, uses a fine-tuned RoBERTa-WWM model to automate the extraction and standardization of unstructured ESG information, forming a computable feature matrix; it concurrently generates subjective and objective weights through the Analytic Hierarchy Process (AHP) and an improved entropy weighting method; introduces game theory mechanisms to solve for Nash equilibrium with the goal of minimizing the sum of squared deviations, obtaining optimal combined weights; based on this, it designs industry dynamic adjustment factors to achieve adaptive calibration of ESG evaluation across industries; ultimately outputting a standardized ESG comprehensive score in the range of [0,1] and mapping it to financial risk control levels. Experimental results show that the proposed system outperforms single weighting methods in weight consistency, score stability, and anti-interference, reducing ESG governance costs for small and medium-sized enterprises, improving evaluation accuracy, and minimizing human interference, providing a lightweight, scalable, and traceable technical path for the micro-implementation of green finance.