ESG Rating Prediction and Management Integrating LightGBM Feature Filtering Algorithm

In response to the low efficiency of current environmental, social, and corporate governance rating prediction management methods in China, a stacked model theoretical framework integrating light gradient boosting machine feature filtering algorithm is proposed and applied to environmental, social, and corporate governance rating prediction management. By improving the feature importance of the output results of the stacked model base learner, the feature imbalance in the model is optimized. Then, a weighted strategy is designed to adjust the fitting performance of the model. The results showed that the improved stacked model increased accuracy by 8.3% and 4.98% respectively compared with the other two models in the self-built dataset. The fitting accuracy of the improved stacked model was 92.32%. The results indicated that the designed method optimized the input layer of the stacked model and enhanced the generalization ability. The weighted optimization strategy of the designed base learner can improve the model fitting effect and enhance the efficiency of stacked models in predicting and managing environmental, social, and corporate governance ratings. The proposed prediction method has positive application significance in environmental, social, and corporate governance rating prediction.