Corporate environmental controversy prediction model verification and application in climate change: Performance improvement based on stacking method

초록

The ability to predict corporate environmental controversies is essential for effective and sustainable business management. However, existing research on corporate environmental performance measurement and controversy prediction remains limited by inconsistent evaluation criteria, insufficient integration of financial and nonfinancial data, and inadequate management of missing values in corporate datasets. In this study, we develop and validate predictive models for environmental controversies by integrating missing value imputation with machine learning approaches, utilizing both financial and environmental data from 1,112 leading U.S. and European companies from 2010 to 2020. First, we demonstrate that applying missing value imputation significantly enhances predictive performance by increasing the reliability of the underlying data, resulting in an improvement of approximately 25% in the area under the precision-recall curve compared to previous studies. Second, models incorporating environmental data (either independently or in combination with financial data) demonstrated significantly superior predictive performance over financial data-only models, with extreme gradient boosting and random forest achieving peak precision (0.83) and area under the receiver operating characteristic curve (0.95) using integrated data. Third, feature importance analysis reveals that variables such as Scope 3 greenhouse gas emissions, waste, hazardous materials, and operating income consistently rank among the most influential predictors, underscoring the necessity for integrated management of key environmental and financial variables to effectively predict and prevent environmental controversies. These findings highlight the importance of integrated data-driven risk management strategies for companies facing increasing climate-related scrutiny and regulatory pressure. The proposed model provides actionable insights for corporate governance and investor decision-making, enhances transparency and accountability, and offers a robust framework for sustainable business operations amid the evolving landscape of environmental, social, and governance disclosure and environmental risk management.

키워드