基于朴素贝叶斯分类的上市公司财务异常侦测研究

发布时间：2018-01-30 06:01

本文关键词： 财务异常朴素贝叶斯分类识别　出处：《吉林大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,伴随着信息技术的广泛应用,企业财务工作方式也在不断的进行改变。财务工作由最初的手工记账,转变为如今信息系统的应用;由简单的财务核算不断的向着更深层次的管理型财务发展,并最终转变成战略型财务。企业的记账方式和财务思维正在逐渐的改变,企业的信息披露也越来越透明化。但无论如何改变,企业的财务状况,一直以来都受到监管部门、政府机构、外部投资者、企业内部经营者的关注。企业财务状况的微小异常对于企业会计假设的持续经营、货币计量会带来重大的影响,同时可能给与企业相关联的各方带来巨大的损失,因此研究企业的财务异常十分重要。引起企业财务异常的原因包含很多种,有因为企业经营出现问题的,有企业出于特定目的进行财务造假的,但不管是什么原因,企业的财务异常最终会体现在财务指标的波动上。本文选取我国2003-2015年间受到处罚的沪深A股上市企业的财务指标作为研究样本,利用SPSS中的显著性检验,对财务异常样本与非异常样本进行方差和均值的检验,确定财务异常与非异常样本存在明显差异的指标共有32个。通过因子分析对所列出的特征属性进行提取,提取出对论文研究具有主要影响的特征因子,形成11个新的财务指标因子进行接下来的研究。通过SPSS将连续型数据离散化,利用基于扫描个案的等百分位将连续型的指标数据转换为离散型指标。通过朴素贝叶斯分类的方法,利用训练样本确定分类器的特征属性的概率值,然后将测试样本应用于形成的分类器,来验证分类器的分类精度,最后本文建立的模型的整体准确率为77.92%,误判率为22.08%,误拒率为12.34%,误识率为9.7%。本文利用朴素贝叶斯的方法研究上市公司的财务异常,通过研究发现企业的财务异常往往受到行业的影响,制造业、房地产业等行业的财务更容易发生异常,近几年发展比较迅速的信息技术业、软件和信息技术业企业的财务异常现象也较为明显。通过分析在所选取的指标中偿债能力、产权结构指标对企业的财务异常具有较强的指示作用,多数企业的财务状况都存在连续多年的异常。
[Abstract]:In recent years, with the extensive application of information technology, the financial work mode of enterprises is constantly changing. The financial work has changed from manual bookkeeping to the application of information system. From simple financial accounting to a deeper level of management financial development, and finally into strategic finance. The accounting method and financial thinking of enterprises are gradually changing. Corporate disclosure is also becoming more and more transparent. However, the financial situation of enterprises has always been subject to regulatory authorities, government agencies, external investors. The concern of the enterprise's internal managers and the slight abnormal financial situation of the enterprise will have a significant impact on the continuous operation of the accounting assumptions of the enterprise and the monetary measurement. At the same time, it may bring huge losses to the parties associated with the enterprise, so it is very important to study the financial anomalies of enterprises. The causes of the financial anomalies of enterprises include many kinds of reasons, some of which are caused by the problems in the management of enterprises. There are enterprises for specific purposes of financial fraud, but no matter what the reason. The financial anomalies of enterprises will eventually be reflected in the fluctuations of financial indicators. This paper selects the financial indicators of Shanghai and Shenzhen A-share listed companies which were punished from 2003 to 2015 as the research samples. Using the significance test in SPSS, the variance and mean value of the financial abnormal sample and the non-abnormal sample are tested. There are 32 indexes to determine the significant difference between the financial anomaly and the non-abnormal sample. The feature attributes are extracted by factor analysis to extract the characteristic factors which have the main influence on the research of this paper. Form 11 new financial index factors to carry on the following research. Discretization of continuous data through SPSS. The continuous index data is transformed into discrete index by equal percentile based on scanning case, and the probability value of feature attribute of classifier is determined by training sample through naive Bayesian classification method. Then the test samples are applied to the formed classifier to verify the classification accuracy of the classifier. Finally, the overall accuracy of the model is 77.92, and the error rate is 22.08%. The false rejection rate is 12.34 and the false recognition rate is 9.7.The paper uses naive Bayes method to study the financial anomalies of listed companies, and finds that the financial anomalies of enterprises are often affected by the industry. Manufacturing, real estate and other industries are more prone to financial anomalies, in recent years, the relatively rapid development of information technology industry. The financial anomalies of software and information technology enterprises are also obvious. Through the analysis of the solvency of selected indicators, the property rights structure indicators have a strong indication of the financial anomalies of enterprises. The financial situation of most enterprises has been abnormal for many years.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F275;F832.51

【参考文献】