Research On The Construction Of Financial Statement Fraud Knowledge Base And The Identification Of Financial Statement Fraudulent Behavior

The financial statement fraud has seriously affected the stability of the social economy and capital market.At the same time,it has brought huge economic losses to capital market participants and affected investors' confidence in the capital market.One of the key factors in the development of the capital market is that the financial statement of listed companies needs to be authentic and complete.Therefore,how to effectively identify the fraudulent behavior of financial statement of listed companies is an important issue for regulators.With the continuous development of big data and artificial intelligence technology,technologies such as data mining,machine learning,natural language processing and semantic analysis are widely used in the financial field,including stock market value investment,intelligent customer service and intelligent risk management.As an important regulator of listed companies,the supervision ability of capital market regulators is one of the key factors determining the success of capital market supervision and listed company supervision.The application of big data and artificial intelligence technology can effectively enhance the regulatory capabilities of regulators.The use of artificial intelligence and other related technologies has provided strong technical support for the discovery and utilization of financial statement fraud knowledge of listed companies,enabling capital market regulators to obtain more efficient and intelligent supervision capabilities.The research of this paper aims to use artificial intelligence to improve the intelligent level of supervision of capital market regulators,and to provide a new method for listed companies to identify fraudulent behaviors in financial reporting.Based on this,this study uses machine learning,deep learning,natural language processing and knowledge base method to discover the financial statement fraud knowledge of listed companies,and builds a financial report fraud ontology knowledge base based on financial indicators,which realizes the financial report fraud based ontology knowledge base.The financial statement fraud knowledge discovery,the financial statement fraud emotional semantic knowledge base was constructed,and the financial statement fraud behavior recognition based on the financial report fraud emotional semantic knowledge base was realized.(1)To sort out the domestic and international research status of fraudulent behavior identification of financial statement of listed companies,analyze and discuss the research status,and point out the existing research problems.The concept of financial statement fraud of listed companies was defined,and the connotation of financial statement fraud was clarified.It sorts out the theory and knowledge discovery theory of financial statement fraud,and expounds the connotation of financial statement fraud knowledge discovery.It sorts out the theory of knowledge base and knowledge base construction methods.It lays a theoretical foundation for the identification of fraudulent behavior in financial statement of listed companies.(2)Proposing the selection of financial indicators for financial statement fraud detection of listed companies.In order to reduce the data dimension and improve the efficiency of financial report fraud detection based on financial indicators,this study combs the existing feature selection algorithm,and proposes a feature selection algorithm based on extra tree to select financial indicators for the subsequent financial statement fraud detection ontology library.The build provides the foundation.(3)Constructing the fraudulent ontology of financial statement of listed companies.This study first combs the ontology related concepts and construction methods and ontology construction tools.According to the financial statement fraud behavior identification task requirements and the ontology design concept,according to the financial index characteristics selection results,the financial report fraud entity of the listed company is constructed.(4)Realizing financial statement fraud knowledge discovery and financial statement fraud behavior identification based on financial statement fraud ontology knowledge base.This study combs the inference engine and decision tree rules to obtain related concepts and methods.This study uses the C4.5 decision tree algorithm to obtain the financial statement fraud detection rules,and uses the SWRL rule language to semantically describe the fraud detection rules and construct a financial statement fraud detection rule base.The financial statement fraud rule base and the financial report fraud ontology library are combined to construct a financial report fraud ontology knowledge base for financial statement fraud knowledge discovery,and financial statement fraud detection based on financial indicators.(5)Propose the construction of financial statement fraud sentiment semantic knowledge base based on financial statement text data,financial domain sentiment vocabulary,deep learning and Word Net semantic knowledge base.This research combs the research status and construction methods of the existing semantic knowledge base construction,and proposes a financial statement fraud vocabulary construction method based on text,financial sentiment seed vocabulary and Word2 vec.Then,using the Word Net semantic knowledge base,the semantic expansion of the financial statement fraud vocabulary based on synonymous and antisense relations is realized,and the financial statement fraud emotional semantic knowledge base is constructed.(6)Implementation of financial statement fraud detection based on financial statement fraud sentiment semantic knowledge base.This study uses the constructed financial statement fraud sentiment semantic knowledge base to semantically mark the financial statement text,uses Word2 vec to express the text based on distributed word vector,and finally uses SVM to realize the financial statment fraud behavior recognition based on text data.