Financial Fraud Prediction in Chinese Growth Enterprise Board Listed Companies

In recent years, there have been frequent incidents of financial fraud in China’s A-share GEM-listed companies. In order to effectively detect instances of financial fraud, this study focuses on 980 listed companies in China’s Growth Enterprise Market (GEM) and utilizes the GWO+XGBoost algorithm to develop a predictive model for identifying such fraudulent activities. This study incorporates both financial and non-financial information from the companies. The empirical studies reveal that machine learning-based models such as SVM and XGBoost exhibit superior predictive performance compared to traditional statistical methods, including Naive Bayes and Logistic regression.; The GWO+XGBoost model outperforms other machine learning models in terms of Precision, Recall, F1 and AUC; The incorporation of non-financial information indicators, such as corporate governance and audit information, significantly enhances the predictive accuracy of the model, underscoring the efficacy of non-financial information in providing valuable incremental information content for financial fraud prediction.; The study also employs Shapley’s value method to examine the contribution of characteristic variables in predicting financial fraud. This analysis provides valuable decision-making guidance for auditors, investors, and regulators, helping to reduce information asymmetry in the capital market and enhance resource allocation efficiency.