受试者工作特征曲线及其在金融危机预警研究中的应用

Receiver Operating Characteristic Curve and Its Application for Early-Warning of Financial Crisis

基于受试者工作特征(ROC)曲线的分析框架,梳理了ROC曲线对预警指标分类能力的评估和预警模型预报准确性评估两方面的应用情况。认为ROC曲线及ROC曲线下面积(AUC)在金融危机预警研究中表现出3个优点:ROC曲线可提供结构法而非经验法来选择预警指标;为指标选择与模型评价提供了统一的分析框架;为不同预警模型的等效性检验提供了分析基础。

Based on the basic principle of receiver operating characteristic （ROC） curve, this paper presented that ROC curve was used for evaluating the classification ability of early-warning indicators and the forecasting accuracy of early-warning systems of financial crises. ROC curve and the area under the curve （AUC） analysis displayed three advantages. First, it proposed a structural method rather than experiential method to choose the predicting indicators. Second, it provided a unified analytical framework of indicator selection and model evaluation. Third, it provided an analytical basis to test the equivalence of different early-warning model.