Research On The Identification Of Financial Fraud In Listed Companies Based On XGBoost Algorith

Capital market is an important part of modern financial market system,which plays the role of allocating market resources,discovering prices and raising funds.China has always attached great importance to guiding the sound development of the capital market.Whether it is the revision of the securities law in recent years or the implementation of the comprehensive registration system,it shows the determination of the Chinese government to vigorously develop the capital market.However,the financial fraud of listed companies is still common,which will not only damage the interests of investors,accumulate market risks,but also restrict the development of the capital market.In today’s increasingly fierce fight against fraud and anti-fraud,the pain point of securities regulatory agencies is that they can not take into account every listed company,and the supervision technology and supervision methods need to be improved.Therefore,this paper builds a financial fraud identification model based on XGBoost algorithm,and conducts a performance test of the model,and then puts forward an effective model method for financial fraud identification of listed companies,which has important reference significance for securities regulators and investors.This paper selects the financial statement data of China’s A-share market from 2016 to 2021,selects 52 financial and non-financial indicators based on the motivation theory of financial fraud and common characteristics of financial fraud,and constructs A financial fraud identification and early warning indicator system.A total of 493 samples from 288 enterprises punished for financial fraud in six years were selected,and the control samples were matched according to the ratio of1:1.XGBoost algorithm was used to train the model,and the performance of the recognizer was tested according to the different needs of securities regulators and investors.At the same time,through the model to select a number of indicators that are important to identify financial statement fraud,and analyze and explain,and then put forward policy recommendations.Finally,the performance of several commonly used machine learning algorithms is compared.The results show that XGBoost algorithm has a good performance in financial fraud identification of listed companies.Specifically: through the verification of a variety of machine learning algorithms,build a financial fraud identification index system including financial indicators and non-financial indicators,which has a strong ability to explain financial fraud and can effectively identify financial fraud.Using XGBoost algorithm for model training has the advantages of high training efficiency and excellent model performance after training.Through the characteristic importance chart,it is found that the five non-financial indicators of the number of litigation cases,the type of audit opinions,the concentration of equity,stock fluctuation and turnover rate play an important role in identifying the financial fraud of listed companies.Based on the empirical analysis results,the following suggestions are put forward: first,pay attention to the important role of enterprise litigation case indicators in identifying enterprise financial fraud,pay attention to the litigation situation of enterprises,send inquiries to enterprises in time,and understand the impact of litigation situation on production and operation;Second,improve the role and status of audit institutions,make them issue audit reports on an independent basis,and find the problems of Listed Companies in time;Third,solve the financing difficulties of enterprises and broaden financing channels for enterprises.