个人信用混合两阶段评估方法研究
|
摘要
随着中国经济的快速发展,各种个人消费贷款的规模迅速扩大。但是,由于目前国内商业银行对零售业务的风险管理水平较低,管理手段与方法相对落后,缺乏一套有效的个人信用评估方法,这严重阻碍了个人消费信贷业务的发展。因此,开发出一套适合于中国特征的,能够有效降低信用风险的个人信用评估方法,对社会经济的发展具有十分重要的意义。
本文运用国内外银行最经常使用的判别分析、Logistic回归、人工神经网络3种方法对收集的数据资料进行处理,并且在对结果进行分析比较的基础上,建立一种新的信用评估方法——个人信用混合两阶段评估方法。即将个人信用评估模型的建立分成两个阶段,在第一阶段通过对判别分析、Logistic回归、人工神经网络3种方法的单独应用所得结果的分析和比较,得到人工神经网络模型的拟合精度最优,Logistic回归预测的稳健性最优的结论,作为第二阶段建立二者混合模型的基础。第二阶段利用Logistic分析建立的模型输出预测结果---概率P,并将P作为新增变量和其他自变量一起作为神经网络的输入,利用先验知识对人工神经网络加以提示,来优化神经网络。由此得到了预测稳健性得到改善的更有效的信用评估模型。这种个人信用混合两阶段评估方法可以有效降低商业银行的信用风险,更好地实现银行利润最大化的目标。
With the rapid development of Chinese economy, the scale of various personal consume expands quickly. But, because of the low risk management level over the retail trade from the interior commercial banks, relatively backward management means and methods, lack of an effective personal credit evaluation method, all severely hindered the development of credit business of personal consume. Therefore, it is very important for the development of social economy to develop an evaluation method of personal credit, which is suitable for the Chinese character and can effectively lower the credit risk.
The paper uses the three methods that are judge and analysis, logistic return and artificial nerve network, which are often used in banks home and abroad, to handle the collected data and form a mixed two-stage evaluation method of personal credit based on the result from the analysis and comparison of the data, that is to divide the personal credit evaluation model into two stages. At the first stage, such conclusions are made that the imitation and synthesis accuracy of artificial nerve network and the predict stability of logistic return are the best through the analysis and comparison of the results from the separate application of the three methods: judge and analysis, Logistic return and artificial nerve network, which is the basement of mixture model of the two in the second stage. The second stage uses the output based on the Logistic analysis model to predict the result---Rate P, puts Rate P as a new variable into the nerve network with other independent variable and gives directions to the artificial nerve network to optimize nerve network. At last we get a more effective credit evaluation model from the improvement of model of imitation and analysis accuracy and predict stability. This research on the mixed two- stage evaluation method of personal credit can reach the goal, that is to effectively lower the credit risk of commercial banks and realize maximize of the bank profits.
本文运用国内外银行最经常使用的判别分析、Logistic回归、人工神经网络3种方法对收集的数据资料进行处理,并且在对结果进行分析比较的基础上,建立一种新的信用评估方法——个人信用混合两阶段评估方法。即将个人信用评估模型的建立分成两个阶段,在第一阶段通过对判别分析、Logistic回归、人工神经网络3种方法的单独应用所得结果的分析和比较,得到人工神经网络模型的拟合精度最优,Logistic回归预测的稳健性最优的结论,作为第二阶段建立二者混合模型的基础。第二阶段利用Logistic分析建立的模型输出预测结果---概率P,并将P作为新增变量和其他自变量一起作为神经网络的输入,利用先验知识对人工神经网络加以提示,来优化神经网络。由此得到了预测稳健性得到改善的更有效的信用评估模型。这种个人信用混合两阶段评估方法可以有效降低商业银行的信用风险,更好地实现银行利润最大化的目标。
With the rapid development of Chinese economy, the scale of various personal consume expands quickly. But, because of the low risk management level over the retail trade from the interior commercial banks, relatively backward management means and methods, lack of an effective personal credit evaluation method, all severely hindered the development of credit business of personal consume. Therefore, it is very important for the development of social economy to develop an evaluation method of personal credit, which is suitable for the Chinese character and can effectively lower the credit risk.
The paper uses the three methods that are judge and analysis, logistic return and artificial nerve network, which are often used in banks home and abroad, to handle the collected data and form a mixed two-stage evaluation method of personal credit based on the result from the analysis and comparison of the data, that is to divide the personal credit evaluation model into two stages. At the first stage, such conclusions are made that the imitation and synthesis accuracy of artificial nerve network and the predict stability of logistic return are the best through the analysis and comparison of the results from the separate application of the three methods: judge and analysis, Logistic return and artificial nerve network, which is the basement of mixture model of the two in the second stage. The second stage uses the output based on the Logistic analysis model to predict the result---Rate P, puts Rate P as a new variable into the nerve network with other independent variable and gives directions to the artificial nerve network to optimize nerve network. At last we get a more effective credit evaluation model from the improvement of model of imitation and analysis accuracy and predict stability. This research on the mixed two- stage evaluation method of personal credit can reach the goal, that is to effectively lower the credit risk of commercial banks and realize maximize of the bank profits.
引文
1 John B.Caouettee, Edwar. Altman, Paul Narayanan. Managing Credit Risk: The Next Great Financial Challenge. Beijing: China Machine Press, 2001.4
2 世界银行.新兴市场经济中的商业银行.北京:中国财政经济出版社,1994.9
3 Pietro Penza, Vipul K. Bansal. Measuring Market Risk with Value at Risk. Penguin,1999.10
4 张维,李玉霜.商业银行信用风险评估综述.管理科学学报,1998.1(3):20~26
5 Anthony Saunders. Credit Risk Measurement: New Approaches to Value at Risk and Other Paradigms.London, 2001.4
6 吴晶妹.资信评估.北京:中国审计出版社,2001.1
7 王春峰等.商业银行信用风险评估及其实证研究.管理科学学报,1998.1,第 1 卷第 1 期:68~72
8 姜明辉,王雅林等.k_近邻判别分析法在个人信用评估中的应用.数量经济技术经济研究.2004.2:143~147
9 张玲.信用风险评估方法发展趋势.预测,2000.4 第 4 期:72~75
10 Basle Committee on Banking Supervision. Credit Risk Modelling: Current Practices and Applications.Basle, 1999, (http:/'www.bis.org/)
11 范彩霞,蒋玉明. 模糊 ISODATA 聚类的 BP 神经网络及其应用.计算机工程与应用.2005.35:226~228.刘中华.银行贷款信用风险管理的发展与完善.上海金融,1998 年第 1 期:41~42
12 王朝勇 .基于 LS_SVM 的个人信用评估 .吉林工程技术师范学院学报.2005.12:5~8
13 刘太安,刘欣颖,张琳.基于 LSVM 的个人信用评估.山东科技大学学报.2005.12:33~35
14 N. Capon. Credit Scoring Systems: a Critical Analysis. Journal of Marketing. 1982, (46): 82~91
15 姜明辉,姜磊,王雅林. 线性判别式分析在个人信用评估中的应用.管理科学.2003.2:53~55
16 徐 少 锋 . FISHER 判 别 分 析 在 个 人 信 用 评 估 中 的 应 用 . 统 计 与 决策.2006.1:133~135
17 An G. The effect of adding noise during backpropagation training on a generalization performance. Neural Computation,1996(8):643~671
18 石庆焱、靳云汇.个人信用评分的主要模型与方法综述.统计研究,2003.8:36~39.28 石庆焱.多种个人信用评分模型在中国应用的比较研究.统计研究.2004.6:43~47
19 王春峰.金融市场风险管理,天津:天津大学出版社,2001.2
20 方洪全,曾勇.银行信用风险评估方法实证研究及比较分析.金融研究2004.1:62~69
21 Risenbeis, R. A Problems in applying discriminant analysis in credit scoring models. Journal of Bankinn and Finace, 2, 205~219
22 施锡铨,邹新月.典型判别分析在企业信用风险评估中的应用.财经研究.2001.10:53~57
23 邹新月,李汉通.运用典型多元判别分析法评估上市公司信用风险.统计与决策.2001.7:21~22
24 方 洪 全 , 曾 勇 . 运 用 多 元 判 别 法 评 估 企 业 信 用 风 险 的 实 例 . 预测.2004.4:65~68
25 李建云,邱菀华.核 FISHER 判别分析方法评估消费者信用风险.系统工程理论方法应用.2004.12:548~556
26 管七梅 . 我国农林牧渔业短期贷款企业违约判别实证分析 . 金融论坛.2006.1:14~19
27 傅强,李永涛. Logistic 模型在上市公司信用风险评价中的应用.重庆建筑大学学报.2005.10:113~116
28 傅强,李永涛.基于 Logistic 模型的上市公司信用风险评价.华东经济管理. 2005.9:95~98
29 徐少锋,王延臣.个人信用评估中的 Logistic 模型.天津轻工业学院学报.2003.12:46~49
30 石 庆 焱 . 多 种 个 人 信 用 评 分 模 型 在 中 国 应 用 的 比 较 研 究 . 统 计 研究.2004.6:43~47
31 卞建华,瞿坦.人工神经网络在商业银行个人信用等级评估与预测中的应用.中国金融电脑.2003.7:77~78
32 胡望斌,朱东华,汪雪锋.商业银行个人信用风险等级评估与预测.商业时代?理论 2005.9:52~53
33 王莉,郑兆瑞,郝记秀.BP 神经网络在信用风险评估中的应用.太原理工大学学报.2005.3:216~219
34 张道宏,张璇,尹成果. 基于 BP 神经网络的个人信用评估模型.情报杂志.2006.3:68~70
35 王春峰,万海晖,张维.金融市场风险测量总体框架.国际金融研究,1998 年第 9 期
36 蒲立丛. 财务分析法——个人消费信用贷款风险评估的探索. 新金融. 2002, (1):21~22
37 Elizabeth Mays. Credit Risk Modeling: Design and Application. Fitzroy Dearborn Publishers, 1998:14
38 Linda Allen, Gayle DeLong, Anthony Saunders. Issues in the Credit Risk Modeling of Retail Markets. Journal of Banking & Finance. 2004, (28): 728
39 孙东升. 商业银行个人金融服务系列讲座之一:个人信用分析. 农村金融研究. 2001,(3):59~61
40 Su-Lin Pang,Yan-Ming Wang,Yuan-Huai Bai. Credit scoring model based on neural network. Machine Learning and Cybernetics, 2002.(4): :1742 – 1746
41 RipleyBD. Neural networks and related methods for classification. J.R.Statist Soc.B,1994,56:409～456
42 David west. Neural network credit scoring models. Cpmputers & operations Research. 2000,(27):1131～1152
43 Pietro Penza, Vipul K. Bansal. Measuring Market Risk with Value at Risk. Penguin,1999.10
44 Oldrich Alfons Vasicek. Credit Valuation. KMV Corporation, New York, March 22,1984
45 石庆焱.一个基于神经网络——Logistic 回归的混合两阶段个人信用评分模型研究.统计研究 2005.5:45~49
46 王 静,王延清,何德权.基于多层前馈神经网络的个人信用评分模型.经济师.2004.12:20~21
47 王晶晶,肖秋霞.个人信用让人工神径网络以 BP 算法评估.系统工程.2005.3:43~45
48 Barron A R.Approximation and estimation bounds for artificial neural networks.Machine Learning,1994(14):115~133
49 Geman S,Bienenstock E and Doursat R. Neural Networks and Bias/VarianceDilemma. Neural Computation,1992(4):1~58
50 Bishop C M. Training with noise is equivalent to Tikhonov regularization. Neural Computation,1995(7):108~116
51 Schmidt W F,Raudys S,Kraaijveld M A,Skurikhina M and Duin R P W. Initialization, Back-Propagation and Generalization of Feed-Forward Classifiers. in Proc.IEEE Int.Conf.Neural Netwoks,1993:598~604
52 Kolen J F and Pollack B J. Back Propagation is Sensitive to Initial Conditions. NIPS 3, 1991:860~867
53 Abu-Mostafa Y S. Learning from hints in neural networks. Journal of Complexity, 1990(6):192~198
54 Abu-Mostafa Y S. A method of learning from hints. NIPS 5, 1993:73~80
55 Wang C and Venkatesh S S. Optimal Stopping and Effective Machine Complexity in Learning.NIPS 6, 1994:303~310
56 Wang C and Venkatesh S S. Temporal Dynamics of Generalization in Neural Networks.NIPS 7, 1995:263~270
2 世界银行.新兴市场经济中的商业银行.北京:中国财政经济出版社,1994.9
3 Pietro Penza, Vipul K. Bansal. Measuring Market Risk with Value at Risk. Penguin,1999.10
4 张维,李玉霜.商业银行信用风险评估综述.管理科学学报,1998.1(3):20~26
5 Anthony Saunders. Credit Risk Measurement: New Approaches to Value at Risk and Other Paradigms.London, 2001.4
6 吴晶妹.资信评估.北京:中国审计出版社,2001.1
7 王春峰等.商业银行信用风险评估及其实证研究.管理科学学报,1998.1,第 1 卷第 1 期:68~72
8 姜明辉,王雅林等.k_近邻判别分析法在个人信用评估中的应用.数量经济技术经济研究.2004.2:143~147
9 张玲.信用风险评估方法发展趋势.预测,2000.4 第 4 期:72~75
10 Basle Committee on Banking Supervision. Credit Risk Modelling: Current Practices and Applications.Basle, 1999, (http:/'www.bis.org/)
11 范彩霞,蒋玉明. 模糊 ISODATA 聚类的 BP 神经网络及其应用.计算机工程与应用.2005.35:226~228.刘中华.银行贷款信用风险管理的发展与完善.上海金融,1998 年第 1 期:41~42
12 王朝勇 .基于 LS_SVM 的个人信用评估 .吉林工程技术师范学院学报.2005.12:5~8
13 刘太安,刘欣颖,张琳.基于 LSVM 的个人信用评估.山东科技大学学报.2005.12:33~35
14 N. Capon. Credit Scoring Systems: a Critical Analysis. Journal of Marketing. 1982, (46): 82~91
15 姜明辉,姜磊,王雅林. 线性判别式分析在个人信用评估中的应用.管理科学.2003.2:53~55
16 徐 少 锋 . FISHER 判 别 分 析 在 个 人 信 用 评 估 中 的 应 用 . 统 计 与 决策.2006.1:133~135
17 An G. The effect of adding noise during backpropagation training on a generalization performance. Neural Computation,1996(8):643~671
18 石庆焱、靳云汇.个人信用评分的主要模型与方法综述.统计研究,2003.8:36~39.28 石庆焱.多种个人信用评分模型在中国应用的比较研究.统计研究.2004.6:43~47
19 王春峰.金融市场风险管理,天津:天津大学出版社,2001.2
20 方洪全,曾勇.银行信用风险评估方法实证研究及比较分析.金融研究2004.1:62~69
21 Risenbeis, R. A Problems in applying discriminant analysis in credit scoring models. Journal of Bankinn and Finace, 2, 205~219
22 施锡铨,邹新月.典型判别分析在企业信用风险评估中的应用.财经研究.2001.10:53~57
23 邹新月,李汉通.运用典型多元判别分析法评估上市公司信用风险.统计与决策.2001.7:21~22
24 方 洪 全 , 曾 勇 . 运 用 多 元 判 别 法 评 估 企 业 信 用 风 险 的 实 例 . 预测.2004.4:65~68
25 李建云,邱菀华.核 FISHER 判别分析方法评估消费者信用风险.系统工程理论方法应用.2004.12:548~556
26 管七梅 . 我国农林牧渔业短期贷款企业违约判别实证分析 . 金融论坛.2006.1:14~19
27 傅强,李永涛. Logistic 模型在上市公司信用风险评价中的应用.重庆建筑大学学报.2005.10:113~116
28 傅强,李永涛.基于 Logistic 模型的上市公司信用风险评价.华东经济管理. 2005.9:95~98
29 徐少锋,王延臣.个人信用评估中的 Logistic 模型.天津轻工业学院学报.2003.12:46~49
30 石 庆 焱 . 多 种 个 人 信 用 评 分 模 型 在 中 国 应 用 的 比 较 研 究 . 统 计 研究.2004.6:43~47
31 卞建华,瞿坦.人工神经网络在商业银行个人信用等级评估与预测中的应用.中国金融电脑.2003.7:77~78
32 胡望斌,朱东华,汪雪锋.商业银行个人信用风险等级评估与预测.商业时代?理论 2005.9:52~53
33 王莉,郑兆瑞,郝记秀.BP 神经网络在信用风险评估中的应用.太原理工大学学报.2005.3:216~219
34 张道宏,张璇,尹成果. 基于 BP 神经网络的个人信用评估模型.情报杂志.2006.3:68~70
35 王春峰,万海晖,张维.金融市场风险测量总体框架.国际金融研究,1998 年第 9 期
36 蒲立丛. 财务分析法——个人消费信用贷款风险评估的探索. 新金融. 2002, (1):21~22
37 Elizabeth Mays. Credit Risk Modeling: Design and Application. Fitzroy Dearborn Publishers, 1998:14
38 Linda Allen, Gayle DeLong, Anthony Saunders. Issues in the Credit Risk Modeling of Retail Markets. Journal of Banking & Finance. 2004, (28): 728
39 孙东升. 商业银行个人金融服务系列讲座之一:个人信用分析. 农村金融研究. 2001,(3):59~61
40 Su-Lin Pang,Yan-Ming Wang,Yuan-Huai Bai. Credit scoring model based on neural network. Machine Learning and Cybernetics, 2002.(4): :1742 – 1746
41 RipleyBD. Neural networks and related methods for classification. J.R.Statist Soc.B,1994,56:409～456
42 David west. Neural network credit scoring models. Cpmputers & operations Research. 2000,(27):1131～1152
43 Pietro Penza, Vipul K. Bansal. Measuring Market Risk with Value at Risk. Penguin,1999.10
44 Oldrich Alfons Vasicek. Credit Valuation. KMV Corporation, New York, March 22,1984
45 石庆焱.一个基于神经网络——Logistic 回归的混合两阶段个人信用评分模型研究.统计研究 2005.5:45~49
46 王 静,王延清,何德权.基于多层前馈神经网络的个人信用评分模型.经济师.2004.12:20~21
47 王晶晶,肖秋霞.个人信用让人工神径网络以 BP 算法评估.系统工程.2005.3:43~45
48 Barron A R.Approximation and estimation bounds for artificial neural networks.Machine Learning,1994(14):115~133
49 Geman S,Bienenstock E and Doursat R. Neural Networks and Bias/VarianceDilemma. Neural Computation,1992(4):1~58
50 Bishop C M. Training with noise is equivalent to Tikhonov regularization. Neural Computation,1995(7):108~116
51 Schmidt W F,Raudys S,Kraaijveld M A,Skurikhina M and Duin R P W. Initialization, Back-Propagation and Generalization of Feed-Forward Classifiers. in Proc.IEEE Int.Conf.Neural Netwoks,1993:598~604
52 Kolen J F and Pollack B J. Back Propagation is Sensitive to Initial Conditions. NIPS 3, 1991:860~867
53 Abu-Mostafa Y S. Learning from hints in neural networks. Journal of Complexity, 1990(6):192~198
54 Abu-Mostafa Y S. A method of learning from hints. NIPS 5, 1993:73~80
55 Wang C and Venkatesh S S. Optimal Stopping and Effective Machine Complexity in Learning.NIPS 6, 1994:303~310
56 Wang C and Venkatesh S S. Temporal Dynamics of Generalization in Neural Networks.NIPS 7, 1995:263~270