基于SAE-BP神经网络的审计风险识别研究——以计算机、通信和其他电子设备制造业行业为例  

Research on Audit Risk Identification Based on SAE-BP Neural Network:A Case Study in the Computer,Communication,and Other Electronic Equipment Manufacturing Industry

作　　者：刘聪粉 张庚珠 LIU Cong-fen;ZHANG Geng-zhu(Business School(School of Management),Northwest University of Political Science and Law,Xi’an 710100,China)

机构地区：[1]西北政法大学商学院(管理学院),陕西西安710100

出　　处：《经济问题》2024年第6期123-128,F0003,共7页On Economic Problems

基　　金：陕西省统战理论与实践课题研究项目“新形势下促进全省民营经济发展壮大研究”(2023HZ1293)。

摘　　要：审计风险的识别和评估是现代风险导向审计的重要内容,为准确地识别审计风险,建立了一套基于SAE-BP神经网络的审计风险识别模型。选取16个指标构成重大错报风险评估模型的输入指标体系,利用SAE算法提取特征,通过机器学习模型BP神经网络分类器进行识别,构建SAE-BP神经网络,并选取135个A股上市公司作为样本进行了实证分析。结果表明:该模型运算速度快,模型平均识别准确率较高,可以达到88.5%,能够对审计风险进行高质量识别,有效提高了审计的效率。The identification and assessment of audit risks are critical components of modern risk-oriented auditing.To accurately identify audit risks,we have developed an audit risk identification model based on the SA E-BP(Stacked Autoencoder-Backpropagation Neural Network).This model employs 16 indicators to constitute an input index system for assessing the risk of significant misstatements.Utilizing the SA E algorithm extracts fea-tures.The BP neural network classifier,a machine learning model,is used for the identification process,culminat-ing in the construction of the SAE-BP neural network.An empirical analysis is conducted using a sample of 135 companies listed on the A--share market.The results demonstrate that the model operates at high speed and boasts a notable average identification accuracy of 88.5%.It can effectively and efficiently identify audit risks,signifi-cantly enhancing the auditing process.

关 键 词：审计风险识别 大数据 稀疏自编码器 神经网络 

分 类 号：F239.1[经济管理—会计学]