本文共六章,第一章绪论,叙述了企业财务预警研究的背景及意义。第二章
相关文献综述,主要对国内外利用神经网络模型进行财务预警的文献进行归纳总
结,为后文的研究奠定基础。第三章概念界定与样本选择,明确界定了本文中财
务危机的概念并确定了研究样本及期间。第四章财务预警指标筛选分析,首先确
定了初选的财务预警指标,在此基础上为了合理确定神经网络模型的输入节点,
提高模型预测的准确性,本文在研究中分别利用统计方法及粗糙集理论对初选的
财务预警指标进行筛选。第五章财务预警模型的构建及实证分析,依据第四章选
取的财务预警指标分别构建神经网络预警模型,并对不同的模型结果进行比较分
析。第六章是结论。
本文以财务预警指标的选取为出发点,分别利用了统计方法及粗糙集理论对
预警指标进行筛选,以此改进神经网络模型的输入节点。通过实证证明了在财务
预警研究中,对预警指标进行适当的预处理会提高神经网络模型预测的准确性。
同时本文通过对比不同的预警指标约简方法,得到利用粗糙集理论对指标进行约
简预测准确性优于传统统计方法,证明了粗糙集理论和神经网络模型相结合的有
效性。
关键词:财务预警;神经网络;主成分分析;粗糙集
Abstract
The enterprise’s finance condition is the reflection of business performance.
Monitoring and forecasting financial statement can assist managers avoiding financial
crisis. This becomes the hot spot of theory and practice research. This paper uses
neural network as financial early warning method as it has strong learning and
fault-tolerant abilities.
There are six chapters in the paper. The first chapter is the introduction. In this
chapter, the paper introduced the background and the significance of the study of
financial crisis warning. The second chapter is literature review. The paper
summarizes and analyzes the literature of neural network both in domestic and
international, which lay the foundation for further study. The third chapter is the
definition and sample selection. In this chapter, the paper clarify the definition of
financial crisis, the sample and duration of the research. The fourth chapter is to build
financial forecasting index system. In this chapter, the paper select the index by the
method of statistics and rough set. The fifth chapter is to construct financial warning
model. The paper build neural network model by using the above warning indexes.
And the last chapter is conclusion.
In order to improve the accuracy of warning, the research uses the statistical
method and rough theory to select the warning indexes. Through the empirical
research, the paper proves that the warning indexes pretreatment will improve the
accuracy of warning. Meanwhile, through the comparison of different methods of
pretreatment, the research also proves that the accuracy of warning by rough theory is
better than it by statistical method.
Key Words: Financial Crisis Warning; Neural Network; Main-composition Analysis;
Rough set Theory