中国海外上市公司的PCA—SVM财务危机预警研究

在支持向量机（Support Vector Machine，SVM）方法的基础上融入主成分分析（Principal Component Analysis，PCA）方法，可构建PCA—SVM财务危机预警模型。以我国海外上市公司为研究对象，运用PCA提取出对财务危机具有显著影响的特征指标，进而通过训练集在不同核函数下对SVM进行训练，最后运用测试集对经过训练得到的SVM财务危机预警模型进行性能验证与评价。实证研究结果表明，PCA．SVM财务危机预警模型在线性、多项式、径向基和sigmoid四种核函数下都具有良好的预测能力，而径向基核函数下的PCA-SVM财务危机预警模型具有更加优越的学习能力与泛化推广能力。

Research on Financial Crisis Warning for Listed Companies Overseas Based on PCA-SVM

This paper integrates the Principal Component Analysis into Support Vector Machine to establish a model of financial crisis warning based on PCA and SVM. Firstly, taking the listed companies overseas as the objects of study, the PCA is used to pick up character indices of financial crisis. And then. training set is adopted to train SVM on account of different kernel functions. At last, testing set is applied to debug and evaluate the capability of SVM financial crisis warning model by trained. The result of investigation illustrates that the financial crisis warning model based on PCA and SVM has fine prediction ability on account of linear, polynomial, radial basis and Sigmoid kernel functions. But the PCA-SVM financial crisis warning model on account of the radical basis function in learning and generalization ability is better.