【作者】 谢润椿；

【导师】 王斌会；

【作者基本信息】 暨南大学 ， 统计学， 2010， 硕士

【摘要】 本文首先以1998-2008年我国A股主板上市公司14042份财务年报数据为基础,对财务风险情况进行了全面的统计分析,探讨了国内上市公司财务风险的主要表现,发现净利润和每股净资产是界定财务风险情况的主要指标。针对上市公司财务变化过程设计了财务演化图,直观地展示上市公司历年财务状况及财务风险变化过程,并在此基础上结合ST制度提出了预警链,全面展示了上市公司各阶段财务状况及其风险变化的上下游关系,且结合历史资料统计了预警链每一环节的情况。在训练样本抽取阶段,采用MKV选样法在722家上市公司1998-2008年年报中选取了393例风险样本和393例配对正常样本进行预警建模,总样本达786例,全面地利用了风险样本信息。在指标体系设计阶段,采用了U检验和因子分析法,阐述了指标体系设计过程。在预警建模阶段,用Logistic回归模型、Fisher判别模型及CHAID决策树模型进行了综合预警分析和比较,结果表明CHAID模型具有良好的预警效果,平均准确率和检出率达到73.7%。并进一步提出了单一型、保守型、风险型预警策略,极大地提高检出率和准确率,为预警链各个环节的预警模型建立提供了示例,最后运用各种预警策略对2010年上市公司的财务风险进行了预测预警。更多还原

【Abstract】 This paper focused on financial risk early-warning for listed company. Base on 14042 annual reports from 1998 to 2008 of main board share A market manufacturing listed company, this paper firstly carried out a comprehensive statistical analysis of the financial risk and discussed the principal representation of it, and found that net profits and PNE are the key indicators of measuring financial risk. Designeing financial-performance-chart to show the changes in the financial situation for listed companies over years intuitively. Creating early-warning-chain which base on ST system of China to display the upstream and downstream relationship amoug every financial period and the changes of risk, and then analyzed every links of the early-warning-chain. In the sampling section, this paper developed a new methor MKV and collected 393 risk samples accompany with 393 normal ones from annual reports from 1998 to 2008 of 722 listed companys, which taked greate advantage of the information of risk samples. U test and Factor analysis are applied for the purpose of building an appropriate index system for early-warning. Then useing Logistic regression, Fisher discriminant analysis and CHAID decision tree to form the early-warning model respectively. The result showed that CHAID has good effects in average detection rate and average correct rate which both were 73.7%. Futherly, this paper developed 3 types of early-warning strategies which called single type, risk type and conservative type, greatly improving the detection rate and correct rate, and then used these strategies to make a forcast of the financial risk of listed companys in 2010.更多还原