论文部分内容阅读
在财务困境预测中,如何从大量备选指标中筛选出预警指标是一个重要环节。为了更有效地设计财务困境预测模型,本文将平均影响值方法应用于SVM回归来进行变量筛选,首先对训练集数据用SVM进行训练,然后分别增减每一自变量的10%来进行仿真,对两个仿真结果的差值按样本数平均,得出平均影响值;最后对各个自变量的平均影响值按绝对值大小排序,从而进行变量筛选。实证结果表明,该方法能够以较少的特征变量实现较高的分类精度,是切实有效的。
In the prediction of financial distress, how to select early warning indicators from a large number of alternative indicators is an important link. In order to design the financial distress prediction model more effectively, we apply the average influence value method to the SVM regression for variable selection. First we train the training set data using SVM, and then increase or decrease 10% of each independent variables to simulate, The difference between the two simulation results is averaged by the number of samples to obtain the average influence value. Finally, the average influence value of each independent variable is sorted according to the absolute value, so as to perform variable selection. The empirical results show that the method can achieve higher classification accuracy with fewer feature variables and is effective.