Research On The Detection Of Financial Reporting Distortion Of Manufacturing Listed Companies Based On Integrated Models

The problem of financial report distortion of listed companies is basically accompanied by the emergence of modern capital markets.Listed companies distort financial data and whitewash financial reports in order to obtain illegitimate benefits or achieve illegitimate purposes.With the development of the capital market,their methods become more diverse,complex,and difficult to regulate,which has a very negative impact on stakeholders in the capital market.However,traditional regulatory methods have significant loopholes in both efficiency and effectiveness,and the detection and early warning of financial report distortion of listed companies has become a major international challenge.Therefore,it is of great theoretical significance and practical value to apply the technology and methods of industrial engineering for process anomaly detection to securities regulation,identify abnormal financial reports,and use big data and artificial intelligence technology to conduct intelligent detection of financial fraud of listed companies.In order to construct an integrated model with excellent performance and high detection accuracy,this article first sorts out and summarizes existing literature.Combining with the actual situation of China’s capital market,the annual financial statement data of manufacturing listed companies is used as a research sample,and combined with the principles of ensemble learning algorithms,a Stacking based financial report distortion detection model is constructed.In the process of building the model,after balancing the positive and negative sample data with the undersampling method,XGBoost,K nearest neighbor,random forest and support vector machine are selected as the primary classifiers of integrated learning;Using a logistic regression model as a secondary classifier for ensemble learning;Finally,the Stacking ensemble learning technology was used to integrate the above algorithms and construct a financial report distortion detection model based on the Stacking ensemble algorithm.It was confirmed that the Stacking ensemble method improved in all four measurement indicators compared to a single model,proving the effectiveness of the Stacking ensemble.Among them,the improvement in precision was the most significant,with an increase of 7.22% compared to the best single model XGBoost,from 74.390% to 79.759%;The recall rate has increased from 61.128% to 63.455%;The Fvalue has also significantly increased,from 65.857% to 70.387%;The accuracy has increased from 80.000% to 81.764%.