Exploring Random Forest Modeling of Accounting Disclosure Quality and ESG Performance in the Face of Big Data Era

• Published in: Applied mathematics and nonlinear sciences Vol. 9; no. 1
• Main Authors: Zhang, Rongkun, Yang, Zi, Liu, Xinxin
• Format: Journal Article
• Language: English
• Published: Beirut Sciendo 01.01.2024; De Gruyter Poland
• Subjects: 00A71; Accounting disclosure; ESG performance; Fixed effects model; Random forest model; TOPSIS method
• ISSN: 2444-8656; 2444-8656
• DOI: 10.2478/amns-2024-3273

This paper combines the theoretical knowledge of the current research on accounting information disclosure and ESG performance, selects seven first-level indicators, which are truthfulness, completeness, materiality, relevance, timeliness, comparability, and comprehensibility, and then subdivides these seven first-level indicators to obtain 22 second-level indicators, thus forming the evaluation index system of accounting information disclosure quality. Based on the high flexibility and robustness of the Random Forest (RF) model, it is proposed to use the Random Forest (RF) model to calculate the weights of the evaluation indicators. The evaluation index system’s comprehensive evaluation results are calculated using the TOPSIS method immediately. The comprehensive evaluation results of accounting information disclosure quality are taken as the explanatory variables of the fixed-effects model, and the data of the explanatory variables and control variables originated from a certain data system, and the fixed-effects model is chosen to clarify the relationship between the quality of accounting information disclosure and ESG performance. The regression coefficient for accounting disclosure quality is 0.0349, and there is a significant positive correlation between accounting disclosure quality and ESG performance at the 1% level. When the quantitative value of corporate accounting disclosure increases by 1 unit, the quantitative value of ESG performance improves by 0.0349 units.