基于HMM-GARCH模型的VaR方法及其在农业股市的应用 VaR Method Based on HMM-GARCH Model and Its Application in Agricultural Stock Market

文章利用隐马尔科夫模型在状态划分上的优势，将隐马尔科夫模型(HMM)与GARCH模型结合建立HMM-GARCH模型来度量金融资产风险价值(VaR)｡首先对收益率序列建立隐马尔科夫模型，将股票市场分为正常状态与异常状态，用Baum-Welch算法估算模型的参数，再采用Viterbi算法估算收益率序列所对应的隐状态序列，根据隐状态序列将收益率序列分为两类，分别建立HMM-GARCH模型估算VaR｡最后利用GARCH模型､MRS-GARCH模型､HMM-GARCH模型对北大荒股票(600598)数据进行了实证分析，采用Kupiec-失败频率检验法对估计的VaR值进行检验｡实证结果表明，基于HMM-GARCH模型的VaR方法能更好的描绘和预测该股票的风险｡ In this paper, the HMM and GARCH models are used to establish the HMM-GARCH model to measure the financial asset risk value (VaR) by using the advantage of the hidden Markov model in state division. First, the hidden Markov model of the financial asset return sequence is set up. The Baum-Welch algorithm is used to estimate the parameters of the model. Then, the Viterbi algorithm is used to estimate the corresponding hidden state sequence of the return sequence. According to the hidden state sequence, the return sequence is classified in two categories. And the HMM-GARCH model is established to estimate VaR for each state sequence respectively. Finally, the Beidahuang stock (600598) data are analyzed by using the GARCH model, MRS-GARCH model and HMM-GARCH model respectively. The accuracy of the VaR is tested by Kupiec failure frequency method. The results show that the HMM-GARCH model can better describe and predict the risk of the stock.