Research On Discriminating Technique Of Accounting Fraud

Accounting fraud of the listed company is a universal phenomenon on today's stock market of the whole world, which not only causes heavy losses to the investor, but also does great harm to the reasonable distribution of the resources and the healthy development of the stock market. Accounting fraud is increasingly prevailing and the losses caused by it is more and more heavy, while anti-fraud practices appear a serious lag instead. The KPMG (1998) survey found that over one third of fraud cases were discovered by accident and that only 4 percent of cases were detected by CPA. Fraud is difficult to detect because it generally is subject to concealment or falsification of accounting records and supporting documents. Recent years some research has tested the applicability of various statistics and modeling technology in the economic and the industrial field. The purpose of this study is attempting to construct a hybrid model for the discrimination of accounting fraud, examining the application effect of the model in the discrimination of accounting fraud, and assessing whether the model can become a tool of complement for the auditors such as CPA, which is based on the reviewing and analyzing of the domestic and foreign methods and models of the discrimination of accounting fraud.In the first chapter, the purpose and foundation of the selection of the thesis is introduced with a brief narration and comment to the present situation of research in the domestic and foreign corresponding field, and the framework and method of the research and the innovation of this paper is brought out as well. Three kinds of technologies concerning accounting fraud are selected and compared in the second chapter, which reported to have good effect in practical use. A brief summary is made after the discovery of the advantages and disadvantages of each technology. The innovation research on discriminating technique of accounting fraud is carried out in the third chapter, wherein a brief introduction is also made to the corresponding theory and concept of the technique of the research as to theoretically prove the applicability and feasibility of the technique, and it is based on the research done in the second chapter. In order to test the validity of the hybrid model brought forward in the paper, some samples and indexes selected are processed with statistical software step by step, and then there is an analysis andsummary of the result, all of which is presented in the fourth chapter. The fifth chapter is the last part of this paper, it introduces the main work done in the article, points out the deficiency in the process of the research and the farther continuing of it. Through theory discussion and software simulation, this article has proven the effectiveness of discriminating accounting fraud using the relationships between entries of financial statement and red flags, and has also proven the applicability of the hybrid model in the discrimination of accounting fraud, which is structured by incorporating fuzzy logic and neural network.