머신러닝 기반 기업성과 예측 모델의 비교 연구: ESG 평가 등급 포함의 영향

· Research topics: This study reminds us of the importance of ESG information and objectively verifies that ESG is a significant factor in predicting corporate performance as a paper proving that ESG ratings are better predicted when included in predicting corporate performance. · Research background: Most corporate performance studies have used ROA as a proxy variable and focused only on regression analysis models, and there have been few attempts on other research methodologies. Therefore, this study aims to develop a corporate financial performance prediction model using machine learning techniques to differentiate it from existing methods. · Differences from prior research: Most previous ESG studies have shown that ESG activities have a significant impact on financial performance using regression analysis models. However, due to data collection restrictions, evaluation of corporate financial performance has relied primarily on the financial factors of financial statements. Therefore, this study aims to develop an ROA prediction machine learning model that considers ESG ratings, a representative non-financial factor. In particular, it is intended to prove that the performance of the corporate performance prediction model is better when the ESG evaluation rating is included. · Research method: In this study, the entire industry was sampled and long-term data of 10 years was used to predict corporate performance. In addition, the classifier method was used among machine learning analysis methods, and more reliable research results were presented by presenting the analysis results of two classifiers (Extra Trees, Random Forest) together. · Research results: In conclusion, these results were consistently derived from Accuracy analysis results, Precision analysis results, Recall analysis results, F1-score analysis results, and AUC analysis results. Therefore, when predicting ROA, which is a measure of corporate performance, it can be seen that predicting including ESG evaluation grade information can predict more accurately than predicting ROA without including ESG evaluation grade information. · Contribution points and expected effects: By proposing a new research methodology that utilizes machine learning in corporate performance research, it provides useful predictive model design guidelines for subsequent researchers. In practice, it provides investors with investment strategies that take into account ESG ratings and meaningful information for companies to recognize the importance of ESG management.