基于谱风险度量的支持向量机理论及在银行信贷中的应用

Support Vector Machine Theory Based on Spectral Risk Measurement and Its Application in Bank Credit

根据一致风险度量构建支持向量机(Support Vector Machine, SVM)的方法,建立了基于谱风险度量的SVM模型。利用优化问题的一阶必要条件确定了所建模型的不确定集的表示。接着,利用某银行金融贷款借贷数据,对新建模型进行了测试。并与基于CVaR的SVM模型、不同核函数下的传统SVM模型、进行对比。实验结果表明,使用基于谱风险度量的SVM模型预测准确度高达82.93%,预测精度相比于基于CVaR的SVM模型要高出3.26%;与传统的SVM模型相比要高出5.49%.证明了基于谱风险度量的SVM模型在金融贷款预测情况下的优越性和高效性。

According to the method of constructing Support Vector Machine(SVM) based on coherent risk measurement,an SVM model based on spectral risk measurement is established.The representation of the uncertainty set of the model is determined by using the first-order necessary conditions of the optimization problem.Then,the new model is tested by using the loan data of a bank.Moreover,it is compared with the SVM model based on CVaR,the traditional SVM model under different kernel functions.The experimental results show that the prediction accuracy of SVM model based on spectral risk measurement is 82.93%,which is 3.26% higher than that of SVM model based on CVaR and 5.49% higher than that of traditional SVM mode.Those results prove the superiority and efficiency of the SVM model based on spectral risk measurement in the financial loan prediction.