AI-Driven ESG Scoring Model for Sustainable Investment in Digital Markets

Abstract

Sustainable investment has become a critical focus area, with Environmental, Social, and Governance (ESG) criteria gaining significant importance in financial decision-making. Traditional ESG evaluation methods often struggle with accuracy, transparency, and consistency, which can hinder effective sustainable investment analysis. This study proposed an AI-driven framework designed to enhance the accuracy of ESG score predictions, utilizing the Slime Mould Algorithm (SMA) for feature selection and Random Forest Regressor (RFR) for model training. The framework optimizes feature selection by identifying the most relevant ESG factors, improving prediction accuracy and reducing model complexity. Key preprocessing steps, including missing value imputation, duplicate removal, outlier handling, and data normalization, are incorporated to ensure data quality and reliability. Experimental results demonstrate improved predictive performance, with the proposed model achieving a Root Mean Square Error (RMSE) of 0.3649, Mean Absolute Error (MAE) of 0.2538, and R² of 0.9912, outperforming existing approaches. These results confirm the effectiveness of the proposed framework for ESG score prediction and highlight its applicability as a decision-support tool for sustainable investment in digital market environments. By providing reliable and transparent ESG score predictions, the proposed model supports data-driven sustainable investment analysis. Future work may explore real-time ESG data integration, model scalability, and explainable Artificial Intelligence (AI) techniques to further enhance ESG assessment across diverse sectors.