• •

基于Twin-SVM的多分形金融市场风险的智能预警研究

• 出版日期:2018-02-15 发布日期:2018-02-25

Intelligent Early Warning for Multifractal Risk of Financial Market based on Twin-SVM

Wang Peng&Huang Xun

• Online:2018-02-15 Published:2018-02-25

Abstract: Based on 5-min high-frequency transaction data of CSI 300 during 11 years, this paper proposes an approach that defines the normal and concern states of the financial market based on multifractal feature. It introduces a new artificial intelligence model of support vector machine (SVM), namely Twin-SVM and carries out a research on early warning for the risk of financial market with the multifractal feature. The empirical result illustrates as follows: (1) the price volatility in Chinese emerging financial market does significant multifractal feature; (2) the normal and concern states defined by multifractal feature parameters not only are exact, but also have present statistical test significance and clear realistic significance; (3) compared with traditional SVM and BP neural network (NN), Twin-SVM has higher prediction accuracy and better prediction stability, that is, Twin-SVM can effectively solve imbalanced sample problem.