| Zhou Weihua,Zhai Xiaofeng,Tan Haowei.Research on Financial Frauds Prediction Model of Chinese Public Companies with XGBoost[J].The Journal of quantitative and technical economics,2022,(7):176-196 |
| 基于XGBoost的上市公司财务舞弊预测模型研究 |
| Research on Financial Frauds Prediction Model of Chinese Public Companies with XGBoost |
| |
| DOI： |
| 中文关键词: XGBoost 机器学习 财务舞弊 预测模型 |
| 英文关键词: XGBoost Machine Learning Financial Frauds Prediction Model |
| 基金项目: |
|
| Hits: |
| Download times: |
| 中文摘要: |
| 研究目标：探讨如何利用大数据和机器学习方法对上市公司财务数据和非财务数据进行分析和挖掘，并应用于上市公司财务舞弊识别和预测。研究方法：提出一种基于机器学习方法的上市公司财务舞弊预测模型Xscore，对上市公司财务舞弊进行预测。研究发现：Xscore模型能够提高模型预测的准确率，在准确率、召回率、AUC指标、KS值、PSI稳定性等方面均优于Fscore模型和Cscore模型，更适合我国上市公司财务舞弊预测。研究创新：基于2000~2020年中国上市公司数据集为观测样本，通过Benford定律、LOF局部异常法、IF无监督学习法，解决了机器学习应用于财务舞弊识别研究时普遍面临的灰色样本问题，甄选兼具领域特性和统计特征的特征变量；首次将XGBoost集成学习方法应用到上市公司财务舞弊预测分析中，有效提高了上市公司财务舞弊准确率。研究价值：本文将XGBoost集成学习方法引入上市公司财务舞弊识别领域，有助于促进人工智能、机器学习在会计学中的研究与应用，为促进上市公司披露高质量的财务信息和维护资本市场秩序提供参考。 |
| 英文摘要: |
| Research Objectives: To explore how to use big data and machine learning methods to analyze and mine financial and non-financial data of listed companies, and apply them to the identification and prediction of financial fraud of listed companies Research Methods: A machine learning method-based financial fraud prediction model Xscore is proposed to predict financial fraud of listed companies Research Findings: Xscore model can improve the accuracy of model prediction, and outperforms Fscore model and Cscore model in terms of accuracy, recall, AUC index, KS value and capture rate, which is more suitable for financial fraud prediction of listed companies in China Research Innovations: Based on the data set of Chinese listed companies from 2000 to 2020 as the observation sample, we solve the gray sample problem commonly faced when machine learning is applied to financial fraud identification research by Benford's law, LOF local anomaly method, IF unsupervised learning method, and select feature variables with both domain characteristics and statistical features; for the first time, we apply XGBoost integrated learning method The XGBoost integrated learning method is applied to the analysis of financial fraud prediction of listed companies for the first time, which effectively improves the accuracy of financial fraud of listed companiesResearch Value: This paper introduces XGBoost integrated learning method into the field of financial fraud identification of listed companies, which helps to promote the research and application of artificial intelligence and machine learning in accounting, and provides reference for promoting the disclosure of high-quality financial information of listed companies and maintaining the order of capital market |
|
View Full Text
View/Add Comment Download reader |
| Close |
|
|
|
