Research On Financial Fraud Identification Of Listed Manufacturing Enterprises Based On Classification Algorithm

Manufacturing is the foundation for building a country and building a strong country.It is the main body of the real economy and the national economy.The financial industry should improve and enhance its support and service to the manufacturing industry on the premise of preventing risks.China has taken a series of measures(such as improving the financial support mechanism,perfecting the capital market and increasing the proportion of direct financing,etc.)to support the transformation and upgrading of manufacturing industry.However,due to the imperfect operation policy of the capital market,financial fraud events occur all the time.These financial fraud cases not only lead to the loss of trust and confidence of investors for the capital market,but also damage the healthy development of listed enterprises,and will affect the orderly operation of the economic market.Therefore,this article,based on the data of manufacturing listed companies,build a comprehensive and scientific index system and analysis of several machine learning classification model,identify the key indicators that have significant impact on the identification of financial fraud,to improve the classification effect of recognition model,which is of great significance to maintain the stability of the financial market and the high-quality development of listed manufacturing enterprises.Firstly,based on the general table of illegal information published by CSMAR database,this paper screened out the listed manufacturing enterprises punished by China Securities Regulatory Commission from 2010 to 2021 for the behavior of fictitious profits,falsely listed assets and false record(misleading statement),and finally obtained 196 fraudulent enterprises.According to the principles of the same industry,same year,market value proximity,etc.,non-fraudulent enterprises were matched 1:1 to obtain the research sample group of this paper.Secondly,the identification indicators are selected from the two aspects of financial information and non-financial information.47 financial indicators are selected from the six aspects of solvency,ratio structure,operating ability,profitability,development ability and per share indicators.From the perspective of corporate governance mechanisms selection nine non-financial indicators,a total of 56 identification index preliminarily established index system.Significance test was carried out on the index system,and41 variables that passed significance test were recorded as the initial index set.PCA and AEnet index sets were obtained by dimensionality reduction of the initial index set by principal component analysis and adaptive elastic network method.Finally,logistic model,random forest model and Ada Boost model were established for the initial index set,PCA index set and AEnet index set respectively.The conclusions are as follows:1.Compared with the model before dimensionality reduction,logistic model,random forest model and Ada Boost model have the best recognition effect on AEnet index set,and the recognition accuracy is more than 74%,while the recognition effect on PAC index set is the worst.This indicates that the principal component dimension reduction method does not improve the identification effect of the model,while the adaptive elastic net method effectively improves the identification effect of the model.2.The empirical results show that the classification effect of Ada Boost model is better than the other two models,and the AEnet-Ada Boost model obtained by combining Ada Boost algorithm with adaptive elastic network method has the best recognition effect,with recall rate and accuracy of76.27% and 80.51%,respectively.Compared with before dimension reduction,it increased 3.39%and 11.02% respectively.3.Variables such as financial expense ratio,cash and cash equivalents turnover ratio,retained earnings per share,inventory-to-income ratio,earnings per share,working capital ratio,cash ratio and owner’s equity ratio have a significant effect on identifying financial fraud.Finally,based on the above conclusions and the motivation theory of financial fraud,in view of the financial fraud problem of listed manufacturing enterprises,effective suggestions are put forward from the three aspects of reducing opportunity factors,curbing greed and need factors and increasing exposure costs.