不良贷款有无回收判别:一类可选变量的支持向量机方法

针对不良贷款有无回收判别问题中属性变量数目多、示性变量比例高的特点,提出了一类可选择变量的支持向量机方法进行判别预测.该方法一是将逐步回归的支持向量机思想应用在模型变量的选择上,二是将线性逐步回归的结果作为模型选择变量的初始状态,解决了传统支持向量机只能使用固定变量的问题. 实证结果显示,该方法不仅提高了样本外预测的正确率, 而且具有很好的稳健性.

Recovery discrimination for nonperforming loan: A support vector machine method with variable selection

To overcome the difficult of too many indicating variables in the analysis of nonperforming loan recovery, this paper proposes a new SVM method which can select variables. Firstly, the stepwise SVM is employed in the selection of model structure. Secondly, the results of linear stepwise regression are used as the initial states of model selection. Empirical results show that the method not only achieves high accurate in out-sample prediction, but also has a high stable performance.