基于RU-SMOTE-SVM的金融市场极端风险预警研究

Research on Extreme Risk Warning for Financial Market Based on RU-SMOTE-SVM

本文以上证综指和深证成指为研究对象,将随机欠采样(RU)、合成少数类过采样(SMOTE)与传统支持向量机(SVM)相结合,提出了一种改进的SVM模型——RU-SMOTE-SVM模型来预测我国金融市场极端风险,并与传统SVM、SMOTE-SVM、RU-SMOTE-NN和RU-SMOTE-DT进行比较。实证结果表明,RU-SMOTE-SVM既优于传统SVM模型,又比SMOTE-SVM具有更高的预测精度,同时还展示出比RU-SMOTE-NN和RU-SMOTE-DT更为优越的预测性能。

Taking the Shanghai Stock Exchange Composite Index and Shenzhen Stock Exchange Component Index as the objects of research, this paper combines Random Under-Sampling（RU）, Synthetic Minority Over-Sampling Technique （SMOTE） with Support Vector Machine （SVM） to establish an improvement SVM RU-SMOTE-SVM, which is applied to predict the extreme risk in Chinese financial market and compared with conventional SVM, SMOTE-SVM, RU-SMOTE-NN and RU-SMOTE-DT. The result of investigation illustrates that RU-SMOTE-SVM not only outperforms conventional SVM, but also has a higher predictive accuracy than SMOTE-SVM, simultaneously, has a more excellent predictive performance than RU-SMOTE-NN and RU-SMOTE-DT.