论文部分内容阅读
ARMA-GARCH模型进行股票价格收益预测时,只考虑了滞后历史数据所包含的信息,而对于在每个滞后时间点的变化趋势信息却未纳入计算模型进行统一考虑,在一定程度上影响了模型分析时序数据时的泛化能力.本文提出了一种基于微分信息的ARMAD-GARCH模型,在包含传统ARMA-GARCH模型对因变量的滞后值以及残差滞后值进行线性回归的基础之上,又在条件均值方程中增加了因变量滞后值的近似微分信息,用以融合股票价格变化趋势信息,提高预测模型对于价格演变方向的判别能力.通过对于不同市场综合股指收益率数据的实证研究表明,ARMADGAR,CH模型在数据除噪,趋势判别以及预测精确度等方面均优于一般的ARMA-GARCH模型.
ARMA-GARCH model only takes into account the information contained in the lagged history data when it comes to the stock price return forecasting. However, the change trend information at each lag time point is not considered in the calculation model and affects the model to a certain extent This paper presents an ARMAD-GARCH model based on differential information. Based on the linear regression of the lagged value of the dependent variable and the residual lag value of the traditional ARMA-GARCH model, The approximate differential information of the dependent variable’s lag value is added to the conditional mean equation to fuse the information of the change trend of the stock price and improve the ability of the predictive model to discriminate the direction of the price evolution.According to the empirical research on the yield data of the comprehensive stock index in different markets, The ARMADGAR, CH model is better than the general ARMA-GARCH model in data denoising, trend identification and prediction accuracy.