基于人工智能方法的金审工程研究

发布时间：2018-04-20 12:00

本文选题：审计 + 审计流程管理　；参考：《天津大学》2007年博士论文

【摘要】： 随着信息技术的迅猛发展,全球性信息化迅速到来,经济领域高科技成果的层出不穷,给审计信息化进程带来一定的压力和动力。本项研究以人工智能理论与技术、审计理论技术为研究共同基础,采用了理论分析和实证检验相结合的研究方法,重点研究了基于人工智能方法的金审工程问题,较为完备的研究了智能化审计总体系统、智能化内部控制评价系统、内部控制评价中的定性仿真以及神经网络在审计智能中的应用。主要研究内容和创新成果如下: (1)首次把智能理论技术与审计理论技术进行了融合。计算机审计是在信息化环境下的一门新的审计学科,是一种崭新的审计方式。构建了计算机审计的基本理论框架,提出了计算机审计的一般模型,主要包括计算机环境下的审计对象、审计作业模式、审计基本方式、审计技术方法等。提出了智能化审计系统的总体结构,包括现场审计实施系统、联网审计实施系统、审计办公系统和信息资源库四个部分。分析了数据采集与转换,包括数据采集与分析的基本特征,数据采集智能化,并构建了审计数据采集与转换系统。 (2)提出并设计了内部控制智能评价系统。从知识角度总结了国外、国内内部控制制度和体系。分析内部控制需要解决的问题,设计了内部控制智能评价系统,包括系统的整体结构、系统的知识库、数据库、推理机制等。将内部控制与定性仿真方法相结合,研究了QSIM算法,设计了基于定性推理的内部控制评价系统结构,以及系统处理流程和系统功能结构等概念模型。把内部控制的业务循环进行了简化与抽象,并对业务循环的各系统模块进行了分解,建立定性的成因推理模型。 (3)研究了人工神经网络在审计智能中的应用,以神经网络为工具进行审计中问题线索的发现。分析了数据挖掘、神经网络与审计的关系,借鉴了国外神经网络在审计方面的应用,建立了神经网络方法在审计分析性复核过程应用模式。选定BP神经网络为工具,构建了人工神经网络与财务危机发现模型,进行了数据实验,表明了方法的有效性。建立了基于自组织特征映射的纳税情况审计模型,税种选择为增值税,进行模型实验,结果表明了模型的有效性。
[Abstract]:With the rapid development of information technology and the rapid arrival of global information technology, high-tech achievements in the economic field emerge in endlessly, which brings certain pressure and power to the process of auditing informatization. This research is based on the theory and technology of artificial intelligence and audit theory and technology, and adopts the research method of combining theoretical analysis and empirical test, focusing on the research of golden audit engineering based on artificial intelligence. The intelligent audit system, the intelligent internal control evaluation system, the qualitative simulation of internal control evaluation and the application of neural network in audit intelligence are studied in this paper. The main research contents and innovative results are as follows: For the first time, intelligent theory technology and audit theory technology are combined. Computer audit is a new auditing subject under the information environment, and it is a new auditing method. The basic theoretical framework of computer audit is constructed, and the general model of computer audit is put forward, which mainly includes audit object under computer environment, audit operation mode, audit basic mode, audit technique method and so on. The overall structure of the intelligent audit system is put forward, which includes the field audit implementation system, the network audit implementation system, the audit office system and the information resource database. The data acquisition and conversion, including the basic characteristics of data acquisition and analysis, the intelligence of data acquisition, and the construction of audit data acquisition and conversion system are analyzed. The intelligent evaluation system of internal control is put forward and designed. From the angle of knowledge, this paper summarizes the internal control system and system at home and abroad. The internal control intelligent evaluation system is designed, including the whole structure of the system, the knowledge base of the system, the database, the reasoning mechanism and so on. This paper combines internal control with qualitative simulation, studies the QSIM algorithm, designs the architecture of internal control evaluation system based on qualitative reasoning, as well as the conceptual model of system processing flow and system function structure. The business cycle of internal control is simplified and abstracted, and the system modules of the business cycle are decomposed, and a qualitative reasoning model is established. 3) the application of artificial neural network in audit intelligence is studied. The relationship between data mining, neural network and audit is analyzed. The application mode of neural network method in the process of audit analysis is established by referring to the application of foreign neural network in audit. The BP neural network is selected as a tool to construct an artificial neural network and a financial crisis detection model. The data experiments show that the method is effective. A tax audit model based on self-organizing feature mapping is established. The tax type is selected as value-added tax, and the model experiment is carried out. The results show that the model is effective.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2007
【分类号】：F239.1;F224

【引证文献】