Corporate Carbon Emission Prediction Based on Multi-Source Data Fusion

Accurate forecasting of corporate carbon emissions is imperative for investors, policymakers, and the global environmental agenda. Traditional methods, primarily leveraging structured data, often fail to capture the nuanced, influential factors pertinent to corporate emissions. This paper introduces a new pathway that integrates multi-source data to construct diverse machine learning models for enhanced prediction. We implemented a semi-supervised machine learning framework to analyze CSR reports from Chinese A-share companies, extracting textual features through a specially developed Corporate Green Action Index (CGAI). We conduct a comparative analysis of carbon emissions prediction using five machine learning models. The experimental results show that the fusion of numerical and textual data provides a new prediction accuracy and climate risk assessment model, and the prediction accuracy is significantly improved compared with traditional models.