Abstract: 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.

Key words: big data, Gibbs sampling, stochastic search, SAS99；financial fraud, fraud triangle theory;the framework of feature extraction