个人信用评估集成模型研究

发布时间：2018-06-03 23:14

  本文选题：个人信用评估 + Bagging模型　； 参考：《南京信息工程大学》2016年硕士论文

【摘要】：伴随着金融自由化及经济全球化进程的不断深入,我国银行业面临着前所未有的挑战。信用风险管理的效率直接关系到商业银行的经营和发展,而研究和构建更加科学、有效的个人信用评估方法,是当前信用风险管理研究的重大课题。目前,国内外学者在个人信用评估方面进行了广泛而深入的研究,设计了许多适应特定金融环境的个人信用评估模型。本文首先对国内外文献从信用评估和集成方法两方面进行梳理,论述了信用风险的概念、成因、特征和个人信用评估的概念,并从传统统计法和现代人工智能法两方面介绍了信用风险评估的方法和模型。在此基础上,将Bagging集成分类法和各单一分类器应用于两组数据库,利用德国和日本两组数据库检验模型的分类精度和稳健性。由实证结果看出,Bagging分类模型都具有较高的分类精度和稳健性,相对单一分类模型,它的分类效果都不错。本文还提出了改进的集成分类模型：Bagging-Bagging集成模型,结果显示,Bagging-Bagging模型的分类效果更好,可以创新性的应用于个人信用评估领域。同为集成方法的Adaboost模型,不论是模型对样本的分类精度还是模型的稳健性,都明显不如Bagging-Bagging和Bagging-决策树模型,更加验证了Bagging集成模型的实用性。因此,可以认为,Bagging集成模型相比其他模型,尤其是分类性能较弱的模型,有较强的提升分类性能的作用,更适合应用于个人信用评估领域。
[Abstract]:With the deepening of financial liberalization and economic globalization, China's banking industry is facing unprecedented challenges. The efficiency of credit risk management is directly related to the management and development of commercial banks. At present, scholars at home and abroad have carried out extensive and in-depth research on personal credit assessment, and designed many personal credit evaluation models suitable for specific financial environment. This paper firstly combs the domestic and foreign literature from two aspects of credit evaluation and integration methods, discusses the concept, causes, characteristics and personal credit evaluation of the concept of credit risk. The methods and models of credit risk assessment are introduced from two aspects: traditional statistical method and modern artificial intelligence method. On this basis, the Bagging integrated taxonomy and each single classifier are applied to two groups of databases, and the classification accuracy and robustness of the model are verified by using the German and Japanese databases. The empirical results show that bagging classification models have high classification accuracy and robustness, compared with a single classification model, its classification effect is good. This paper also proposes an improved integrated classification model: Bagging-bagging model. The results show that the Bagging-bagging model has better classification effect and can be applied to personal credit assessment innovatively. Both the classification accuracy of the model and the robustness of the model are obviously lower than those of the Bagging-Bagging and the agginging-decision tree model, which verifies the practicability of the Bagging integration model. Therefore, compared with other models, especially those with weak classification performance, the bagging ensemble model has a stronger function of improving classification performance and is more suitable for personal credit evaluation.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：F224;F203
，

本文编号：1974649