数据挖掘技术在商业银行客户信用风险评估中的应用研究
|
摘要
随着金融体制改革的深入,我国商业银行的经营模式和经营理念发生了很大变化。面对日益开放的金融市场和金融全球化的发展,银行经营环境也在不断变化,竞争日趋激烈。同时由于金融创新的日新月异,商业银行的业务不断多样化,风险正在不断增大。市场风险、流动性风险、利率风险、汇率风险正以更新的形式摆在银行面前,其中,信用风险是商业银行所面临的主要风险。而银行客户信用风险的存在,不仅影响了信贷资质,而且干扰了正常的市场秩序。银行和客户之间的信息不对称是其根源。为了提高我国商业银行的信贷经营管理水平,增强市场竞争能力,运用现代信息技术建立科学的客户信用评估体系是十分必要的。
论文首先阐述了商业银行信用风险的相关理论,通过对影响银的行信用风险因素进行分析,明确了对银行客户进行信用评估的意义,同时介绍了评估客户信用风险的指标以及风险衡量的标准。其次,作者对比了传统信用评估方法和现代信用评估方法的优劣。根据信用风险构建原则及客户评估指标的建设情况,建立了对商业银行企业型客户的信用评估指标体系,并用主成分分析法选取了最具代表性的指标,以此得出企业综合评分来作为其后神经网络模型预测准确度的一个对比参照值。再次,作者以机械设备类上市企业的指标数据作为研究样本,采用改进的BP神经网络构建了客户信用风险评估模型。经过实证分析,结果表明该模型能够比较准确的反映出指标体系与风险度之间的映射关系,达到了预期目标。最后,作者说明了神经网络的优势以及BP算法的不足之处,以及改进的方向。
With the Financial market competition’s fury, the operation mode of commercial bank in our country has changed a lot. Also because of the rapid innovation of Financing, commercial bank business has been diversified, at the same time the risk increases. Market risk、liquidity risk、interest risk and exchange rate risk appeared in a new form. Among them, credit risk is the key point to commercial bank. Client credit risk will not only influence credit aptitude, but also can disturb the market system. The root of this is Information Asymmetry between client and the banks. So, in order to increase the level of credit management, enhanced marketability of commercial banks, it’s very necessary to constitute scientific credit assessment system with modern information technology.
In this paper, first introduce the theory of credit risk, analyzed the generation factor of commercial bank credit risk, explicit the signification of client credit assessment ,meanwhile present assessment index of credit risk and the reference point of it. Then, the writer compared traditional method of assessment and modern method. Constituted assessment index system of corporate client credit risk of commercial bank, based on constitution principles of credit risk , and use principal component analysis to select the most influential factors. After that, the writer use index data of machinery listed companies as the samples choose modified BP neural network model to build up client credit risk assessment model. By positive analysis, the result showed this model can reflect urozzlxoek of index system and risk more precisely, attained the expected goal. At last, the writer explained the advantage of neural network and deficiency, also the direction of further perfection.
论文首先阐述了商业银行信用风险的相关理论,通过对影响银的行信用风险因素进行分析,明确了对银行客户进行信用评估的意义,同时介绍了评估客户信用风险的指标以及风险衡量的标准。其次,作者对比了传统信用评估方法和现代信用评估方法的优劣。根据信用风险构建原则及客户评估指标的建设情况,建立了对商业银行企业型客户的信用评估指标体系,并用主成分分析法选取了最具代表性的指标,以此得出企业综合评分来作为其后神经网络模型预测准确度的一个对比参照值。再次,作者以机械设备类上市企业的指标数据作为研究样本,采用改进的BP神经网络构建了客户信用风险评估模型。经过实证分析,结果表明该模型能够比较准确的反映出指标体系与风险度之间的映射关系,达到了预期目标。最后,作者说明了神经网络的优势以及BP算法的不足之处,以及改进的方向。
With the Financial market competition’s fury, the operation mode of commercial bank in our country has changed a lot. Also because of the rapid innovation of Financing, commercial bank business has been diversified, at the same time the risk increases. Market risk、liquidity risk、interest risk and exchange rate risk appeared in a new form. Among them, credit risk is the key point to commercial bank. Client credit risk will not only influence credit aptitude, but also can disturb the market system. The root of this is Information Asymmetry between client and the banks. So, in order to increase the level of credit management, enhanced marketability of commercial banks, it’s very necessary to constitute scientific credit assessment system with modern information technology.
In this paper, first introduce the theory of credit risk, analyzed the generation factor of commercial bank credit risk, explicit the signification of client credit assessment ,meanwhile present assessment index of credit risk and the reference point of it. Then, the writer compared traditional method of assessment and modern method. Constituted assessment index system of corporate client credit risk of commercial bank, based on constitution principles of credit risk , and use principal component analysis to select the most influential factors. After that, the writer use index data of machinery listed companies as the samples choose modified BP neural network model to build up client credit risk assessment model. By positive analysis, the result showed this model can reflect urozzlxoek of index system and risk more precisely, attained the expected goal. At last, the writer explained the advantage of neural network and deficiency, also the direction of further perfection.
引文
1数据来源:《四大国有商业银行上市:2万亿不良资产是道坎儿》,周菡,《证券时报》,2009‐10‐19。
[1] Altman E I. Discriminate Analypis ajd the Prediction of Corporate Bankruptcy [J], Journal of Finance, 1968, 23(3): 580-620.
[2] Ahtman E I. Assessing Potential Financial ProbleMs for Fires in Brazil[J], NYU Sadmmon Center, 1997, 9: 32-80.
[3] Altman E, Marco G, Corporate Distress Diagnosis: Comparisons Using Linear Discriminant Analysis and Neural Networks, Banking and Finance, 1994, 18: 509-540.
[4] Aziz, A D& Emanuel and G. Lawson, Bank Prediction8 An Invaspigation of Cash Flow Based Mode, [J], Journal of Management Studies, 1988, 10: 76-92.
[5] Becker B G. Using Mineset for Knowledge Discoverx [J]. IEEE ComPuter Graphics and Application, 1997 , 29: 42-51.
[6] Clister Lund. Credit Scoring and Credi4 Control [M]. New York: Oxford University Press, 1993.
[7] Coats P, Fant L., Recognizing financial distress patterns using a neural network tool [M], Financial Management, 1993.
[8]David West. Neural network credit scoring models [J]. Computers & Operations Research, 2000, 27.
[9] Feng Yulin, Sally Maclean. A Data mining approach to the prediction of corporate failure [J]. Knowledge-Based Systems, 2001, 14
[10] Hsieh Nan CHEN. Hybrid Mining Approach in the Design of Credit Scoring Models [J]. Expert Systems with Applications, 2005, 28:123-157.
[11] Jensen Herbert, Using neural networks for credit scoring [J]. Managerial Finance, 1992, 18: 69-84.
[12] Jobst, Norbert J., Zenios, On the simulation of portfolios of interest rate and credit risk sensitive securities [J]. European Journal of Operational Research, 2005,161: 298.
[13] Rosch, Daniel. An empirical comparison of default risk forecasts from alternative credit rating philosophies [J]. International Journal of Forecasting, 2005,21(1): 37.
[14] Til Schuerman. A review of recent books on credit risk [J]. Journal of Applied Econometrics, 2005, 20(1): 123-130/
[15] Yang Hailiang. Ruin theory in a financial corporation model with credit risk [J]. Mathematics and Economics, 2003, 33(1): 135-145.
[1]陈京民,数据仓库与数据挖掘技术[M].电子工业出版社,2002.
[2]陈晓,陈治鸿,企业财务理论、方法及应用[J],投资研究,2000, 2.
[3]陈述云,我国银行信贷风险评估模型研究[J ].统计与决策,1998,9.
[4]陈刚,我国商业银行客户信用评估方法应用研究[J],市场现代化,2006,8.
[5]迟国泰,郑杏果,杨中原.基于主成分分析的国有商业银行竞争力评价研究[J],管理学报,2009,6(2).
[6]邓祖新, SAS系统和数据分析[M],电子工业出版社,2002.
[7]戴相龙、吴念鲁,商业银行经营管理[M],中国金融出版社,1998.
[8]范雷文,储海林,上市公司财务指标的综合评价[M],财金贸易,2000,2.
[9]谷秀军.中国信用评级业:走制度规范之路[N].金融时报,2008-07-11.
[10]郝丽萍,胡欣悦,李丽.利用BP网络模型对商业银行信贷风险进行分析和预测[J].系统工程理论与实践. 2001, 5.
[11]何晓群.现代统计分析方法与应用[M].中国人民大学出版社,1999.
[12]胡守仁.神经网络应用技术[M].国防科技大学出版社,1998.
[13]李丽.商业银行信用风险管理策略[J].农场金融研究,2006,3:34-35.
[14]李林.数据挖掘在商业银行信用风险管理中的应用研究[D].电子科技大学,2005.
[15]李绍伟、宋文萍.我国商业银行信用风险的化解之道[J].金融经济,2006,4:32-34.
[16]李月.论我国商业银行的信用风险管理[J].经济纵横,2005,10:34-36.
[17]李志辉、李萌.我国商业银行信用风险识别模型及其实证研究[J].经济科学,2005,5:61-71.
[18]李志辉.现代信用风险量化度量和管理研究[M].中国金融出版社,2001.
[19]林钧跃.企业信用管理[M].企业管理出版社,2001.
[20]林燕、王建琼、刘晓伟.现代信用风险管理模型比较分析[J].商场现代化,2005,30:259-260.
[21]刘桂英.企业信用管理的效益分析[J].财会月刊,2007,2.
[22]刘世平.商业银行信用风险管理纵横谈[J].中国金融家,2005,2:102-104.
[23]潘琰,程小可.上市公司经营业绩的主成分评价方法[J].会计研究,2000,1.
[24]胥振兴,陈启安.基于BP算法的信用评估模型的应用研究[J].福建电脑,2007
[25]唐春阳、张恒、马若微. KMV模型在商业银行信用风险中的应用[J].财会研究,2006,2:19-22.
[26]田继敏、胡红兵、刘鹏.以内部评级法改进信用风险限额管理[J].农场金融研究,2006,3:22-25.
[27]王春峰,万海晖.组合预测在商业银行信用风险评估中的应用[J].管理工程学报,1999,13: 5-8.
[28]王小平.现代信用风险量化模型及对我国商业银行信用风险管理的借鉴[J].经济师,2006,4:251-252.
[29]王栓琴.当前我国商业银行信用风险及防范对策[J].中国市场,2006,4:142.
[30]薛锋,柯孔林.粗糙集——神经网络系统在商业银行贷款五级分类中的应用[J].系统工程理论与实践,2008,(1).
[31]阎庆民、蔡红艳.中国银行不良资产证券化信用风险评价研究[J].数量经济技术经济研究,2005,8:74-83.
[32]杨保安,季海.基于人工神经网络的商业银行贷款风险预警研究[J].系统工程理论与实践,2001.
[33]杨妙华.谈谈改进我国商业银行信用风险管理工作的对策[J].金融经济,2006,12:142-143.
[34]杨雄胜等,企业综合评价指标体系研究[J].财政研究,1998, 5.
[35]于立勇.商业银行信用风险评估[M].北京大学出版社,2007.
[36]赵晓菊,银行风险管理[M].上海财经大学出版社,1999.
[37]赵晓菊,柳永明.金融机构信用管理[M].中国方正出版社,2004.
[38]赵影.企业信用评级指标的选取[J].国际融资,2009,11:54-55.
[39]张银奎等译,数据挖掘原理[M].机械工业出版社,2003.
[1] Altman E I. Discriminate Analypis ajd the Prediction of Corporate Bankruptcy [J], Journal of Finance, 1968, 23(3): 580-620.
[2] Ahtman E I. Assessing Potential Financial ProbleMs for Fires in Brazil[J], NYU Sadmmon Center, 1997, 9: 32-80.
[3] Altman E, Marco G, Corporate Distress Diagnosis: Comparisons Using Linear Discriminant Analysis and Neural Networks, Banking and Finance, 1994, 18: 509-540.
[4] Aziz, A D& Emanuel and G. Lawson, Bank Prediction8 An Invaspigation of Cash Flow Based Mode, [J], Journal of Management Studies, 1988, 10: 76-92.
[5] Becker B G. Using Mineset for Knowledge Discoverx [J]. IEEE ComPuter Graphics and Application, 1997 , 29: 42-51.
[6] Clister Lund. Credit Scoring and Credi4 Control [M]. New York: Oxford University Press, 1993.
[7] Coats P, Fant L., Recognizing financial distress patterns using a neural network tool [M], Financial Management, 1993.
[8]David West. Neural network credit scoring models [J]. Computers & Operations Research, 2000, 27.
[9] Feng Yulin, Sally Maclean. A Data mining approach to the prediction of corporate failure [J]. Knowledge-Based Systems, 2001, 14
[10] Hsieh Nan CHEN. Hybrid Mining Approach in the Design of Credit Scoring Models [J]. Expert Systems with Applications, 2005, 28:123-157.
[11] Jensen Herbert, Using neural networks for credit scoring [J]. Managerial Finance, 1992, 18: 69-84.
[12] Jobst, Norbert J., Zenios, On the simulation of portfolios of interest rate and credit risk sensitive securities [J]. European Journal of Operational Research, 2005,161: 298.
[13] Rosch, Daniel. An empirical comparison of default risk forecasts from alternative credit rating philosophies [J]. International Journal of Forecasting, 2005,21(1): 37.
[14] Til Schuerman. A review of recent books on credit risk [J]. Journal of Applied Econometrics, 2005, 20(1): 123-130/
[15] Yang Hailiang. Ruin theory in a financial corporation model with credit risk [J]. Mathematics and Economics, 2003, 33(1): 135-145.
[1]陈京民,数据仓库与数据挖掘技术[M].电子工业出版社,2002.
[2]陈晓,陈治鸿,企业财务理论、方法及应用[J],投资研究,2000, 2.
[3]陈述云,我国银行信贷风险评估模型研究[J ].统计与决策,1998,9.
[4]陈刚,我国商业银行客户信用评估方法应用研究[J],市场现代化,2006,8.
[5]迟国泰,郑杏果,杨中原.基于主成分分析的国有商业银行竞争力评价研究[J],管理学报,2009,6(2).
[6]邓祖新, SAS系统和数据分析[M],电子工业出版社,2002.
[7]戴相龙、吴念鲁,商业银行经营管理[M],中国金融出版社,1998.
[8]范雷文,储海林,上市公司财务指标的综合评价[M],财金贸易,2000,2.
[9]谷秀军.中国信用评级业:走制度规范之路[N].金融时报,2008-07-11.
[10]郝丽萍,胡欣悦,李丽.利用BP网络模型对商业银行信贷风险进行分析和预测[J].系统工程理论与实践. 2001, 5.
[11]何晓群.现代统计分析方法与应用[M].中国人民大学出版社,1999.
[12]胡守仁.神经网络应用技术[M].国防科技大学出版社,1998.
[13]李丽.商业银行信用风险管理策略[J].农场金融研究,2006,3:34-35.
[14]李林.数据挖掘在商业银行信用风险管理中的应用研究[D].电子科技大学,2005.
[15]李绍伟、宋文萍.我国商业银行信用风险的化解之道[J].金融经济,2006,4:32-34.
[16]李月.论我国商业银行的信用风险管理[J].经济纵横,2005,10:34-36.
[17]李志辉、李萌.我国商业银行信用风险识别模型及其实证研究[J].经济科学,2005,5:61-71.
[18]李志辉.现代信用风险量化度量和管理研究[M].中国金融出版社,2001.
[19]林钧跃.企业信用管理[M].企业管理出版社,2001.
[20]林燕、王建琼、刘晓伟.现代信用风险管理模型比较分析[J].商场现代化,2005,30:259-260.
[21]刘桂英.企业信用管理的效益分析[J].财会月刊,2007,2.
[22]刘世平.商业银行信用风险管理纵横谈[J].中国金融家,2005,2:102-104.
[23]潘琰,程小可.上市公司经营业绩的主成分评价方法[J].会计研究,2000,1.
[24]胥振兴,陈启安.基于BP算法的信用评估模型的应用研究[J].福建电脑,2007
[25]唐春阳、张恒、马若微. KMV模型在商业银行信用风险中的应用[J].财会研究,2006,2:19-22.
[26]田继敏、胡红兵、刘鹏.以内部评级法改进信用风险限额管理[J].农场金融研究,2006,3:22-25.
[27]王春峰,万海晖.组合预测在商业银行信用风险评估中的应用[J].管理工程学报,1999,13: 5-8.
[28]王小平.现代信用风险量化模型及对我国商业银行信用风险管理的借鉴[J].经济师,2006,4:251-252.
[29]王栓琴.当前我国商业银行信用风险及防范对策[J].中国市场,2006,4:142.
[30]薛锋,柯孔林.粗糙集——神经网络系统在商业银行贷款五级分类中的应用[J].系统工程理论与实践,2008,(1).
[31]阎庆民、蔡红艳.中国银行不良资产证券化信用风险评价研究[J].数量经济技术经济研究,2005,8:74-83.
[32]杨保安,季海.基于人工神经网络的商业银行贷款风险预警研究[J].系统工程理论与实践,2001.
[33]杨妙华.谈谈改进我国商业银行信用风险管理工作的对策[J].金融经济,2006,12:142-143.
[34]杨雄胜等,企业综合评价指标体系研究[J].财政研究,1998, 5.
[35]于立勇.商业银行信用风险评估[M].北京大学出版社,2007.
[36]赵晓菊,银行风险管理[M].上海财经大学出版社,1999.
[37]赵晓菊,柳永明.金融机构信用管理[M].中国方正出版社,2004.
[38]赵影.企业信用评级指标的选取[J].国际融资,2009,11:54-55.
[39]张银奎等译,数据挖掘原理[M].机械工业出版社,2003.