基于灰色神经网络的审计意见预测模型研究

发布时间：2018-07-09 19:36

本文选题：预测 + 审计意见　；参考：《重庆理工大学》2017年硕士论文

【摘要】：上市公司定期公布的财务报告是利益相关者进行决策的依据,注册会计师作为独立的第三方对被审计单位财务报告进行审计出具的鉴证性审计意见,对增强上市公司财务信息的可信性起着至关重要的作用,经审计后的财务报表更可靠,使信息使用者能更加有信心,能减少投资决策失误,注册会计师出具的审计意见类型对公司利益相关者的投资决策行为具有重大影响,因此对上市公司审计意见预测模型的研究具有一定的现实意义。本文构建了基于灰色神经网络的上市公司审计意见预测模型,为保证预测效果,在以往研究大多使用财务指标的基础上,建立了包括财务指标和公司治理、会计师事务所等非财务指标的预测指标体系。选取2013-2015年沪深两市全部A股制造业被出具非标准审计意见的上市公司作为研究样本,根据公司的资产规模、行业以及1:1配比原则,并剔除缺失值之后,最终得到训练组、检验组以及预测组样本;为降低训练的复杂性,保证预测效果,采用邻域粗糙集方法对样本指标进行筛选,在不影响预测效果的情况下,剔除次要指标,以保留相对重要指标;利用BP神经网络方法对训练组样本数据进行训练建模及回判仿真,并利用检验组数据对所建立的模型进行检验;选取预测组样本时点数据,构建基于灰色神经网络方法的审计意见预测模型,该模型综合了灰色预测模型建模所需信息较少及神经网络模型自适应、自学习的优点。研究结果表明,加入非财务指标的综合预测模型的预测准确度高于仅用财务指标建模的预测准确度,构建的基于灰色神经网络方法的审计意见预测模型具有较好的预测能力。
[Abstract]:The financial reports issued regularly by listed companies are the basis for stakeholders to make decisions. Certified public accountants, as independent third parties, audit the financial reports of audited units and issue an authenticated audit opinion. It plays an important role in enhancing the credibility of financial information of listed companies. The audited financial statements are more reliable, so that the information users can have more confidence and can reduce the mistakes in investment decisions. The type of audit opinion issued by CPA has great influence on the investment decision-making behavior of corporate stakeholders, so it is of practical significance to study the prediction model of audit opinion of listed companies. In this paper, a forecasting model of audit opinion of listed companies based on grey neural network is constructed. In order to ensure the forecasting effect, the financial indexes and corporate governance are established on the basis of the financial indicators used in most previous studies. Accounting firms and other non-financial indicators of the forecast index system. From 2013 to 2015, all the A-share manufacturing companies in Shanghai and Shenzhen stock markets were selected as the research samples. According to the company's asset size, industry and 1:1 matching principle, and after eliminating the missing value, the training group was finally obtained. In order to reduce the complexity of training and ensure the prediction effect, the neighborhood rough set method is used to screen the sample index, and the secondary index is eliminated without affecting the prediction effect, so as to retain the relative important index. BP neural network method is used for training modeling and simulation of training group sample data, and the test group data is used to test the established model, and the prediction group sample time point data is selected. An audit opinion prediction model based on grey neural network is constructed. The model combines the advantages of less information needed in grey prediction model and adaptive and self-learning of neural network model. The research results show that the prediction accuracy of the comprehensive forecasting model with non-financial indexes is higher than that of only using the financial indicators, and the prediction model based on the grey neural network method has a better forecasting ability.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F239.4

【参考文献】