Research On Fraud Discrimination Of Financial Statements Based On Machine Learning

Fraud in financial statements will not only make accounting information lose its authenticity,but also make information users make wrong decisions,thus affecting the allocation of market resources and hindering the benign development of the market,and even leading to the economic crisis of the whole society.How to effectively identify fraud in financial statements has always been the focus of academic attention.Based on the above background,this paper first adopts the literature research method to review the relevant literature on financial statement fraud means,financial statement fraud screening model,and financial statement fraud motivation theory.Secondly,this paper uses normative analysis method to analyze the concepts related to fraud in financial statements,the means of fraud in financial statements,the commonly used machine learning model for the screening of fraud in financial statements,and the theoretical basis of fraud in financial statements.Thirdly,this paper adopts the empirical research method,selects and determines the fraud screening indicators of financial statements based on the triangle fraud theory,determines the index weight based on the MLP neural network model,thus constructs the fraud risk index of financial statements,and sets the threshold score according to different risk levels,and describes the fraud risk in detail.Finally,this paper uses the case analysis method to further understand how the fraudulent means of statements are operated in enterprises,and then puts forward the case inspiration.This study has significant significance and value for finding and preventing fraud in financial statements.Fushun Co.,Ltd.currently has the largest special smelting furnace group in China and the highest international technical level,and has undertaken the task of developing and producing new special steel materials of the country for a long time.The special steel products produced by Fushun Co.,Ltd.have played an important role in various fields such as aviation,military and national defense.In July 2019,the CSRC issued a notice of administrative punishment to Fushun Co.,Ltd.,confirming that Fushun Co.,Ltd.has engaged in fraud in financial statements for eight consecutive years and inflated profits by 1.902 billion yuan.The reasons for choosing Fushun Co.,Ltd.as the study case in this paper are as follows: on the one hand,the financial statement fraud of Fushun Co.,Ltd.involves a huge amount of money,lasts for a long time,and has a serious impact.On the other hand,Fushun Co.,Ltd.as a large state-owned enterprise,its fraudulent behavior in financial statements not only caused investors to suffer huge losses,but also made the state-owned assets face the risk of loss.In the aspect of financial statement fraud screening research,this paper does not stay on the selection of indicators,but on the construction of screening model for further exploration.On the one hand,the screening principles of traditional machine learning models are mostly based on "supervised learning".The disadvantage of such models is that due to the uncertainty of non-corrupt firms in the training sample,the applicability of the models is severely weakened.On the other hand,due to practical considerations,it is difficult for us to make an accurate judgment on whether there is fraud in financial statements of enterprises,and enterprises are often not simply classified as fraud and non-fraud.Therefore,what we should consider is the size of fraud risk in financial statements.Indexed field research,this paper,based on the theory of the fraud triangle opportunity,pressure and excuse factors,put forward the ideas,build financial statements fraud index with a continuity index instead of two classification model,the objective to quantify the size of the enterprise financial statements fraud risk,in order to make up the defects of existing research classification rough.In the selection of indicators and the determination of index weight,this paper uses MLP neural network model to replace the traditional Delphi method,which makes it more accurate and objective.The main research results of this paper show that: Firstly,the financial statements of Fushun Co.,Ltd.fraud means mainly divided into two steps.The first step is to directly include the raw materials that should have been carried forward into the inventory account in the form of "returning steel",so as to significantly reduce the cost of the main business and achieve the main purpose of directly increasing the current profit.The second step is to hide the inflated inventory by transferring it to the project under construction in batches,and carrying forward the capitalization of fixed assets at the right time for depreciation.This can not only gradually "digest" the inflated inventory,but also indirectly adjust the current profit through the depreciation of fixed assets.Secondly,the fraud in the financial statements of Fushun Co.,Ltd.has caused serious consequences,causing huge losses to small and medium investors and creditors,resulting in a crisis of confidence for investors and a blow to the confidence of the capital market.At the same time,the financial statement dance of Fushun Co.,Ltd.weakened the usefulness of accounting information decision-making.Thirdly,through the construction of financial statement fraud index,detailed description of the financial statement fraud risk of Fushun Co.,Ltd.: unreasonable shareholding structure,state-owned enterprise nature leads to the lack of checks and balances and the audit institution is not independent provides opportunities for fraud;The purpose and motive of fraud are the continuous deterioration of industry operating conditions and delisting pressure.The ethical distortions of corporate executives become excuses for fraud.Fourthly,from the perspective of fraud triangle theory,this paper puts forward some enlightenment and governance suggestions: improve the ownership structure and internal control system of the company,enhance the independence of the company,so as to curb the occurrence of fraud;By perfecting the diversified delisting system,making reasonable corporate strategic objectives to reduce the pressure of fraud;By strengthening the construction of corporate culture and staff self-cultivation to reduce management fraud excuses.Through the construction of fraud index in financial statements,this paper hopes to provide research ideas on the identification and prevention of fraud in financial statements,and provide reference indicators for researches on related financial statements.