基于ROC曲线的银行危机识别方法的效果分析  

Effect Analysis of the Banking Crisis Identification Method Based on the Roc Curve

作　　者：陶菁[1] 余垠 TAO Jing;TAO Jing(School of Sciences,Shanghai Polytechnic University,Shanghai 201209,China;School of Business,Ningbo University,Zhejiang315211,China)

机构地区：[1]宁波大学商学院,浙江315211 [2]上海第二工业大学理学院,上海201209

出　　处：《上海第二工业大学学报》2017年第4期295-304,共10页Journal of Shanghai Polytechnic University

基　　金：教育部人文社科规划项目(12YJAZH183);上海第二工业大学校学科建设项目(XXKPY1604)资助

摘　　要：利用受试者工作特征(ROC)曲线及曲线下面积,将对信号识别效果的要求转换成3个统计评价标准。应用这些标准,从识别能力和稳定性2个维度、全样本和子样本2个方面测试了基于指标的和基于货币压力指数的2种银行危机定义方法。研究发现,基于指标的危机定义的识别能力好于基于货币市场压力指数的识别能力,而名义利率下的货币压力指数的识别效果好于实际利率下的识别效果。不同的危机识别信号效果随着时间的推移和样本范围的不同则有较大的波动。The requirement for signal recognition effect was transformed into three statistical evaluation standards by using ROC curve and area under the curve.Two methods of banking crisis definition including based on moneymarketpressure and based on indicator were tested by applying these standardsfrom two dimensions of recognition ability and stability and two aspects of the whole sample and the subsamples.It was found that identification ability based on indicator was better than that based on money market pressure,and the identification ability based on moneymarketpressure in the nominal interest ratewas better than that in the real interest rate.Different effect of crisis identification signal had larger fluctuation with change in time and range of sample.

关 键 词：银行危机识别 受试者工作特征曲线 曲线下面积 

分 类 号：F831[经济管理—金融学]