基于VaR的中国股指期货风险实证研究

发布时间：2018-01-02 19:14

本文关键词：基于VaR的中国股指期货风险实证研究　出处：《北方工业大学》2013年硕士论文　论文类型：学位论文

【摘要】：2010年4月16日,我国推出了沪深300股指期货,在其推出以后,沪深300股指期货交易比较活跃,运行也较为平稳,但是在风险的度量及预测方面的研究略显不足。 VaR作为一种度量风险的指标,其应用较为广泛,国内外专家学者对其也有较多的研究。为了测度中国股指期货VaR风险,选择中国股指期货当月连续IFO日对数收益率数据作为研究对象,首先对收益率序列进行了基本的统计分析,结果表明,中国股指期货对数收益率序列具有尖峰厚尾的特征,并具有ARCH效应,因此认为可以对此时间序列数据运用GARCH族模型来克服其异方差性。介绍并选用了GARCH族模型中的GARCH-N,GARCH-t,GARCH-GED,TARCH与EGARCH模型度量了中国股指期货的风险,即其VaR数值,并根据各模型系数讨论了中国股指期货的杠杆效应等风险特征。同时,文章还引入了神经网络中较为新颖的混合密度网络模型(MDN模型)对中国股指期货的风险VaR进行考察。作为神经网络的一支,它特有的学习特性可以很好地拟合时间序列的事变特征,从而具有良好的预测能力。最后将GARCH族模型与混合密度网络模型得到的VaR值进行了准确性检验——Kupiec失败频率检验,发现GARCH-GED、EGARCH与混合密度网络模型可以较为准确地预测中国股指期货的风险VaR值。
[Abstract]:In April 16th 2010, China launched the Shanghai and Shenzhen 300 stock index futures, after its introduction, the Shanghai and Shenzhen 300 stock index futures trading is more active, the operation is more stable. However, the research on risk measurement and prediction is slightly inadequate. As a measure of risk, VaR is widely used, and many experts and scholars at home and abroad have studied it. In order to measure the VaR risk of Chinese stock index futures. This paper selects the data of IFO daily logarithmic rate of return of Chinese stock index futures as the research object. Firstly, the paper carries on the basic statistical analysis to the return rate series, the result shows that. The logarithmic return sequence of Chinese stock index futures has the characteristics of sharp peak and thick tail and ARCH effect. It is considered that the GARCH family model can be used to overcome the heteroscedasticity of this time series data. This paper introduces and selects the GARCH-NU GARCH-T GARCH-GARCH-GEDTARCH model and EGARCH model to measure the risk of Chinese stock index futures. That is, its VaR value, and according to the model coefficients to discuss the leverage effect of Chinese stock index futures and other risk characteristics. At the same time. The paper also introduces a novel hybrid density network model in the neural network) to investigate the risk VaR of Chinese stock index futures as a branch of the neural network. Its special learning characteristic can fit the incident feature of time series well, thus it has good prediction ability. Finally, the VaR values obtained from GARCH family model and hybrid density network model are tested for accuracy-Kupiec failure frequency test, and GARCH-GED is found. EGARCH and mixed density network model can accurately predict the risk VaR value of Chinese stock index futures.
【学位授予单位】：北方工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F832.51

【相似文献】