基于KMV模型和支持向量机的上市公司财务危机预警研究

发布时间：2018-05-09 20:44

本文选题：财务危机预警 + 违约距离　；参考：《西南财经大学》2014年硕士论文

【摘要】：企业所有利益相关者,即企业的所在者,企业的经营者,股东高度关注企业的财务状况。对于企业的所有者来说对企业的财务状况的关注就好比每个人对于自己身体状况的关注一样,营运良好的企业可以为企业的所有者带来良好的回报并且可以扩展融资渠道以及获得更多的支持和便利。对于企业的经营者来说,企业的财务状况良好说明经营者过去的某段时间的成绩良好,经营者可以获得更高的薪酬和业绩记录。相反,若企业的财务出现危机,不仅仅会是使股民望而却步还令对原来的企业的经营者和所有者产生恐慌。通过上市公司财务危机预警的研究,我们首先可以制定一个科学的财务危机预警模型,能帮助公司及早根据预警信号采取相应措施,同时对国家证券监管部门监控上市公司质量和减少市场风险也有重要的指导意义。纵观以往的财务危机预警研究,主要可以分为两大类：财务指标预警模型和信用风险量化模型。财务指标预警模型包括单变量模型,多元线性判别模型,线性概率模型,多元逻辑回归模型。信用风险度量模型主要有：KMV公司的KMV模型、JP摩根的信用度量术模型(ceditmetrics model)、麦肯锡公司的宏观模拟模型(credit portfolio view)、瑞士信贷银行的信用风险附加法模型(cridetrisk+)、死亡率模型(mortality rate)等。本文则要利用经典的KMV模型来计算上市公司的违约距离DD. 本文认为SVM模型在财务预测方面会优于其他模型,不仅仅因为其他方法有着诸多的参数假设,还因为在处理小样本的问题上,SVM有着稳定,能克服“维数灾难”等诸多优点。本文首先对所选取的指标进行了简单的解释,本章选取了反映公司盈利能力,偿债能力,成长能力和营运能力的四大指标,它们分别是：净资产收益率,总资产净利率,销售净利率,销售毛利率,流动比率,速动比率,产权比率,存货周转率,应收账款周转率,总资产周转率,营业收入增长率,净利率增长率,总资产增长率,净资产增长率。选取这些指标的原因有两个,其一是这些指标大体上涵盖了一个企业在财务方面的表现,其二是参考了有些学者的研究。其次对KMV原理进行介绍,通过参数的设定进而求得每家公司的违约距离DD。很多学者再采用KMV模型求违约距离DD时,参数的设定都有所不同,比如在DP=STD+1/2LTD这个公式上,有学者就认为DP-STD+O.93LTD。根据国内的资本市场的情况和很多学者的研究的理论基础上,本文认为不必要调整经典KMV模型的参数,根据经典KMV模型,最后本文利用MATLAB算出了违约距离DD。接着本文简单的介绍了数据包络分析(DEA)的原理,并求出了相应的TE(技术效率,代表是企业的投入产出效率)。这部分比较重要的是如何选取DEA的输入和输出指标,本文则参考了潘洁的研究。基于本文样本较少而指标较多的事实,本章决定对样本所选取的指标进行主成分分析,提炼出样本指标的大部分信息。这部分简单的介绍了主成分分析的思想和数学上的表示并且算出了样本的5个主成分值。最后一部分是介绍了支持向量机的思想并且计算出整个企业财务预警模型的预测效率,通过将DD,TE和5个主成分进行SVM分析,本文得到了预测正确概率高达80%以上的财务预警模型。相对于其他学者做的有关财务预警的SVM模型,本文创新性的加入了KMV模型中的违约距离DD作为上市公司的信用指标,TE作为上市公司的投入产出指标,并且预测也取得了良好的效果。
[Abstract]:All stakeholders, that is, the owner of the enterprise, the operator of the enterprise, and the shareholders, are highly concerned about the financial situation of the enterprise. For the owner of the enterprise, it is like the attention to the state of the body for the financial situation of the enterprise. The good operating enterprise can bring good returns to the owner of the enterprise. And it can expand financing channels and get more support and convenience. For business operators, the financial situation of the enterprise is good to show that the manager has done well in the past some time, and the operator can obtain higher salary and record of performance. The deterrent also caused panic to the operators and owners of the original enterprises.
Through the study of the financial crisis early warning of listed companies, we can first make a scientific financial crisis early warning model, which can help the company to take corresponding measures according to early warning signal, and also have important guiding significance for the state securities regulatory department to monitor the quality of listed companies and reduce the market risk.
The previous financial crisis early-warning research can be divided into two major categories: financial indicators early warning model and credit risk quantification model. The financial indicators early warning model includes a single variable model, multiple linear discriminant model, linear probability model, and multiple logistic regression model.
The credit risk measurement models are mainly: KMV company's KMV model, JP Morgan's credit measurement model (ceditmetrics model), the macro simulation model (credit portfolio view) of the McKinsey Co, the credit risk additional model (cridetrisk+), the mortality model (mortality rate) and so on. The model is used to calculate the default distance of a listed company DD.
This paper considers that the SVM model will be superior to other models in financial forecasting, not only because other methods have many parametric assumptions, but also because of the stability of SVM in dealing with small sample problems, and can overcome many advantages such as "dimensionality disaster".
This article first gives a simple explanation of the selected indicators. This chapter selects four major indicators that reflect the company's profitability, solvency, growth capacity and operating capacity. They are net asset returns, net interest rate, net interest rate, gross profit, liquidity ratio, rate of movement, property ratio, and inventory turnover. The accounts receivable turnover rate, total asset turnover rate, business income growth rate, net interest rate growth rate, total asset growth rate and net asset growth rate are two reasons for selecting these indicators. The first is that these indexes generally cover the financial performance of an enterprise, and the other is a reference to some scholars' research.
Secondly, the KMV principle is introduced, and the default distance of each company is obtained through the parameter setting. DD. many scholars use KMV model to find the default distance DD, and the setting of the parameters is different. For example, on the DP=STD+1/2LTD formula, some scholars believe that DP-STD+O.93LTD. is based on the domestic capital market and many scholars. On the basis of the theoretical research, we think that it is unnecessary to adjust the parameters of the classical KMV model. According to the classical KMV model, we use MATLAB to calculate the default distance DD..
Then this paper briefly introduces the principle of data envelopment analysis (DEA), and finds out the corresponding TE (technical efficiency, the representative is the input-output efficiency of the enterprise). This part is more important to select the input and output index of DEA. This article refer to Pan Jie's research. Based on the fact that the sample is less and the index is more, this chapter decides In this part, we simply introduce the idea and mathematical representation of the principal component analysis and calculate the 5 main components of the sample. The last part introduces the thought of support vector machines and calculates the financial early-warning model of the whole enterprise. Through the SVM analysis of DD, TE and 5 principal components, we get the financial early-warning model with the correct probability of more than 80%.
Compared with other scholars' SVM model on financial early-warning, this paper innovatively joins the default distance DD in the KMV model as the credit index of the listed company, TE as the input and output index of the listed company, and the prediction also has achieved good results.

【学位授予单位】：西南财经大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;F275;F224

【参考文献】