회계정보 기반 ESG 성과 예측에 대한 탐색적 연구 -Ensemble 기법을 중점으로-

This study aims to develop a model for predicting ESG (Environmental, Social, and Governance) performance usingcorporate accounting information. ESG performance is a critical indicator for assessing corporate sustainability andsocial responsibility, increasingly becoming a key evaluation criterion for investors and stakeholders. The study wasconducted in two primary stages. The first stage involved data preprocessing, where 427 indicators of accountinginformation were collected and refined to predict ESG performance. This data was extracted from financial statements,income statements, and cash flow statements, providing a comprehensive reflection of the companies’ financial statusand performance. The second stage focused on analysis, employing various data mining techniques and ensemble methods to developand evaluate the ESG performance prediction models. Data mining techniques are invaluable for identifying patternsand rules within large datasets. In this research, both basic data mining techniques and ensemble methods were utilized. The results indicated that the Random Subspace with DT (J48) method exhibited the highest values in terms of accuracy(0.854), precision (0.877), and F-measure (0.888). Additionally, ensemble methods demonstrated improved predictiveperformance compared to single models. This study suggests that ESG performance prediction models can provide valuable insights for evaluating corporatesustainability and making informed investment decisions. The research underscores the significant role of ESGperformance prediction in ensuring long-term corporate success and sustainability, highlighting the necessity ofpredicting and managing ESG performance. By employing various data mining techniques and ensemble methods, thisstudy contributes to enhancing the accuracy of ESG performance predictions and increasing their practical applicability.