财务欺诈风险特征筛选框架的建立和应用

The Framework for the Risk Feature Extraction Method on Corporate Financial Fraud George

本文从金融科技大数据出发,以人工智能的吉布斯随机搜索(Gibbs Sampling)算法为工具,在大数据框架下建立了针对公司财务欺诈风险的特征因子筛选的一般处理方法与特征提取推断原理,并结合上市公司的财务报表数据进行实证分析,结合从2017年1月到2018年12月证监会对上市公司财务报表信息披露违规的数据样本,筛选出刻画财务欺诈的特征因子并进行了验证测试,支持财务欺诈的识别。本文提出的框架和模型方法可以加强和提升对上市公司财务欺诈风险的识别能力,并实现对公司财务在欺诈方面的探测与预测(Detecting and Predicting)功能。

By employing the Gibbs sampling skill under the Markov Chain Monte Carlo(MCMC), weestablish a general framework for corporate financial fraud detection by using fintech method related big data analysis. In the empirical analysis, based on those event “bad” samples from Chinese A-share listed companies enquired by China Securities Regulatory Commission(CSRC)due to behaviors such as violating(at least potentially violating) the rules of the disclosure during time period from the beginning of year 2017 to the end of year 2018 under the Rule of the Disclosure from CSRC, the analysis for key risk factors which could represent the information for the exposure of financial fraud behavior is conducted byudetecting the difference between their financial reports from others. In general, the feature extraction(or variable selection) from around two hundred related factors of financial reports will be a NP problem because of the diversity of financial ratio indexes. However, in this paper by employing the Gibbs sampling method under MCMC, 8 key factors are extracted which are highly correlated with the behavior of corporate financial fraud. They are: ROE, the growth construction-in-process, the growth of advance payment, interest expense/revenue, investment income/revenue, other income/revenue, other receivables/total assets, andlong term loan/total assets.The key contribution of this paper is that a general framework is established for the extraction of key risk factors which could be used not only to detect the behavior of financial fraud, but also to predict the financial fraud under the supporting of ROC testing numerical results based on more than 3,500 A share listed companies in China.