The Research On BP-LVQ Combined Neural Network Fraud Risk Identification Model

In recent years,fraud scandals of listed companies emerge at home and abroad endlessly,which not only brings investors a huge investment risk and damage,but also a heavy blow to the social public confidence in the accounting profession and the capital market.Therefore,how to effectively identify company fraud becomes the focus of accounting theory circle and accounting practice circle as well as regulatory.The empirical study shows that the effect of fraud recognition model is superior to the fraud case analysis.The perfect fraud identification index and proper identification method help to improve the recognition effect.Now the research in fraud identification indicators is abundant,but the research of fraud identification model is very few.With the continuous development of artificial intelligence technology and the wide application,artificial neural network technology begins to be applied in the areas of fraud identification more and more frequently.Among them,BP neural network and LVQ neural network are more widely used in the field of fraud identification with higher fraud recognition rate.Based on this background,this paper delves into these two kinds of neural network technology,using the same sample test to check these two kinds of model.And then put forward an advanced combined neural network fraud risk identification model based on BP and LVQ.After looking up and organizing related literatures at home and abroad,this paper divides the content into eight chapters.In the second chapter,it simply expounds the six of the most popular management fraud motivation and cause theory,and then combs the related literatures about fraud risk identification index and fraud risk identification methods.This part let us know the research status quo,research results and the existing shortcomings of the fraud risk identification models.Based on the second chapter,I put forward the reason of selecting BP and LVQ neural network technology as the fraud risk identification models.In the third chapter,the characteristics and classification of artificial neural network technology are introduced in detail.And also introduces the structures and operation mechanisms of BP neural network and LVQ neural network.Chapter four mainly describes the selection of research samples and the screening of fraud risk identification index.The paper selects 506 listed companies which have corrupt practices as samples for fraud.In order to match the fraud samples,I selects 506 normal companies according to the principle of Beasley.Based on the literature review,I comb some good recognition effect fraud risk identification index as the initial index system,testing paired samples by T test and principal component analysis to eliminate collinearity problem,finally selected 10 indicators whose recognition effect are best.The fifth chapter tests the fraud risk identification effects of BP and LVQ neural network model,and analyzes the fraud discriminant effects of two kinds of recognition model.Sixth chapter proposes BP—LVQ combined neural network fraud risk identification model based on the analysis of BP and LVQ neural network model’s advantages and disadvantages.It introduces the constructing principle and way of thinking.After using the same sample tests fraud recognition effect of combination model,I find that the combination model of discriminant rate was significantly higher than that of single neural network model.In addition,using 2015 sample data to test the stability of combination model find fraud recognition effect of it is better and stable.Finally,according to the above theoretical analysis and empirical research summarizes the full text,analyzes the deficiencies in the study,and puts forward the future development prospects of fraud risk identification model.