我国上市公司财务舞弊识别模型对比研究

发布时间：2018-05-29 04:08

本文选题：上市公司 + 财务舞弊　；参考：《西北大学》2012年硕士论文

【摘要】：财务舞弊泛世界化,严重阻碍了证券市场公开、公平及公正的发展,成为各国资本市场健康有序成长的屏障。在此背景下,如何有效的识别财务舞弊以净化资本市场的诚信环境成为一个亟待解决的问题。因此,本文的研究重点即建立一套行之有效的财务舞弊识别模型。本文首先阐释国内外财务舞弊识别模型的相关文献,为设计识别模型的识别指标提供基础及依据；第二,明确财务舞弊的概念并对财务舞弊进行经济学理论分析；第三,分析财务舞弊的动机、手段及征兆,并就本文应选取何种技术手段构建识别模型做出说明；第四,以1998-2009年期间的120组舞弊样本和配对的非舞弊样本为研究对象,选取相关的识别指标,建立四种财务舞弊识别模型,包括多元逻辑回归识别模型、BP神经网络识别模型、概率神经网络识别模型和Elman神经网络识别模型；第五,以“期望错误分类成本法”为依据对比四种识别模型,分析得出Elman神经网络技术在财务舞弊识别问题上识别准确率最高；最后,基于上述研究结论,总结全文并提出研究的不足之处。本文的创新点在于为财务舞弊识别引进了一种新的技术手段——Elman神经网络,分析发现因其具有历史回溯性等特点,因此可在庞大的财务数据信息中准确地把握财务舞弊识别规律,对提高监管部门判别财务舞弊的准确性和效率方面都起到了积极作用。
[Abstract]:The universal financial fraud seriously hinders the open, fair and just development of the securities market, and becomes a barrier to the healthy and orderly growth of the capital markets of various countries. In this context, how to effectively identify financial fraud in order to purify the integrity of the capital market environment has become a problem to be solved. Therefore, the focus of this paper is to establish a set of effective financial fraud identification model. This article first explains the domestic and foreign financial fraud identification model related literature, provides the foundation and the basis for the design identification model identification index; second, clarifies the financial fraud concept and carries on the economic theory analysis to the financial fraud; third, This paper analyzes the motivation, means and symptoms of financial fraud, and explains what technical means should be selected in this paper to construct the identification model. Fourthly, 120 groups of fraud samples and matched non-fraud samples from 1998 to 2009 are taken as the research objects. Four kinds of financial fraud identification models are established, including multiple logic regression identification model and BP neural network identification model, probabilistic neural network identification model and Elman neural network identification model. On the basis of "expected error classification cost method", four recognition models are compared, and it is concluded that Elman neural network technology has the highest accuracy in identifying financial fraud. Finally, based on the above research conclusions, Summarize the full text and put forward the deficiencies of the research. The innovation of this paper lies in the introduction of a new technique for the identification of financial fraud-Elman neural network. Therefore, it can accurately grasp the identification law of financial fraud in the huge financial data information, which plays a positive role in improving the accuracy and efficiency of judging financial fraud.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F275;F832.51;F224

【引证文献】