基于神经网络的基金评级结果质量检验

发布时间：2018-06-04 16:58

本文选题：基金评级 + 质量检验　；参考：《浙江工业大学》2013年硕士论文

【摘要】：伴随着经济全球化的深入,证券投资基金,银行次级贷款,股票等资本市场的繁荣,信用评级在我国金融市场中的功能和作用日趋显现。投资者,监管部门对评级机构的依赖程度也逐步提高,因此对理论研究工作者也有了更高的要求。如何检验信用评级的信息含量；如何提高评级的公信力及投资者的信息质量,如何借鉴国外先进经验并结合我国国情,建立科学合理的指标体系来提高评级的准确性等等都成为未来我国信用评级研究的基本问题。信用评级的功能在于减少信息不对称,揭示风险,而要实现这一功能,评级质量的保证至关重要。基于目前此种现状,本文企图建立一种检验评级质量的方法。评级质量的优劣可以从两个方面进行衡量：一是评级观点准确性,即验证评级机构“高级别低违约”的基本假设。二是评级观点的稳定性,只有在基本的信用风险发生改变后才会发生级别变更,由于级别所依赖的基本信用风险变化缓慢,因此级别的变动也不应剧烈。因此,本文也从这两个角度切入来对评级机构的评级结果进行质量检验。本文首先采用自组织神经网络的方法结合描述基金质量特征的指标数据对基金进行等级划分,将划分结果与评级机构的评级结果进行对比,以检验其准确性；其次,基于模糊评价法和熵权法,结合迁移矩阵对评级机构的评级结果的稳定性进行检验,并运用随后两个月的基金业绩表现来对本文检验方法进行检测,得出该检验方法具有良好的预测性和适用性。至此,本文从准确性和稳定性两个角度对评级机构的评级结果进行质量检验,这是本文的切入视角创新;本文采用神经网络的方法进行准确性检验,采用迁移矩阵的方法进行稳定性检验这是技术创新。希望通过本研究方法可以完善我国基金评级质量检验体系,促使评级机构真正发挥评级机构的风险揭示和衡量作用。
[Abstract]:With the deepening of economic globalization and the prosperity of the capital market such as securities investment fund, bank subprime loan and stock market, the function and function of credit rating in our country's financial market is becoming more and more obvious. Investors and regulators are increasingly dependent on rating agencies, so they have higher demands on theoretical researchers. How to test the information content of credit rating, how to improve the credibility of credit rating, how to improve the information quality of investors, how to learn from the advanced experience of foreign countries and how to combine the situation of our country, Establishing a scientific and reasonable index system to improve the accuracy of credit rating has become a basic problem in the future study of credit rating in China. The function of credit rating is to reduce information asymmetry and reveal risk. To realize this function, the guarantee of rating quality is very important. Based on the present situation, this paper attempts to establish a method to test the quality of rating. The quality of ratings can be measured from two aspects: first, the accuracy of rating views, that is, to verify the rating agencies "high-level low default" basic assumptions. Second, the stability of the rating view, only after the change of the basic credit risk will occur level change, because the level of the basic credit risk changes slowly, so the level of change should not be violent. Therefore, this paper also analyzes the quality of rating agencies from these two angles. This paper first uses the self-organizing neural network method to combine the index data describing the fund quality characteristic to carry on the grade division to the fund, carries on the comparison between the division result and the rating agency's rating result, in order to test its accuracy, secondly, Based on fuzzy evaluation method and entropy weight method, combined with migration matrix, the stability of rating agencies' rating results is tested, and the performance of funds in the following two months is used to test the test method of this paper. It is concluded that the test method has good predictability and applicability. So far, this paper carries on the quality test to the rating agency's rating result from two angles of accuracy and stability, which is the innovation of the angle of view of this paper, this article uses the neural network method to carry on the accuracy test, The method of migration matrix is used to test the stability. This is a technological innovation. It is hoped that this research method can perfect the quality inspection system of fund rating in our country, and promote the rating agencies to give full play to the role of risk disclosure and measurement of rating agencies.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP183;F832.51

【参考文献】