Detection of Corporate Environmental Information Disclosure Falsification Based on Support Vector Machine

Abstract

Environmental information disclosure (EID) is an important embodiment of corporate social responsibility. With the awakening of public awareness of environmental protection and the increasing pressure of environmental preservation, enterprises tend to strategically manipulate environmental information for the pursuit of profit, which will consequently lead to environmental information disclosure falsification (EIDF) and disruption of both the market regulatory order and the development of green economy. In this article, support vector machine (SVM) technique is applied to construct the detection model of corporate EIDF. Based on the theory of “public pressure,” the detection indicators will be improved from three aspects: public pressure, corporate governance, and financial indicators. The training set and test set are constructed by combining the manually collected cases of environmental administrative penalties from 2015 to 2019 with the indicator information of nonfinancial listed enterprises in China’s A-share market, and the SVM detection performance is compared with the logistic regression of the benchmark model. To solve the problem of category imbalance, we have introduced the Borderline-SMOTE oversampling technique. Based on the detection results of SVM and Borderline-SMOTE, we find that the Borderline-SMOTE-SVM model has the best detection performance, surpassing the SVM and logistic regression models. These conclusions have constructive policy implications for regulatory agencies, investors, the third-party service sector, enterprises, and government policy-making to achieve high-quality corporate EID.