基于贝叶斯因子模型金融高频波动率预测研究

发布时间：2018-03-03 15:38

本文选题：已实现波动率的预测　切入点：HAR模型　出处：《管理科学学报》2017年08期 　论文类型：期刊论文

【摘要】：构建了包含时变系数和动态方差的贝叶斯HAR潜在因子模型(DMA(DMS)-FAHAR),并对我国金融期货(主要是股指期货和国债期货)的高频已实现波动率进行预测.通过构建贝叶斯动态潜在因子模型提取包含波动率变量、跳跃变量和考虑杠杆效应的符号跳跃变量等预测变量的重要信息.同时,在模型中加入了投机活动变量,以考察市场投机活动对中国金融期货市场波动率预测的影响.预测结果表明,时变贝叶斯潜在因子模型在所有参与比较的预测模型当中具有最优的短期、中期和长期预测效果.同时,具有时变参数和时变预测变量的贝叶斯HAR族模型在很大程度上提高了固定参数HAR族模型的预测能力.在股指期货和国债期货的预测模型中加入投机活动变量可以获得更好的预测效果.
[Abstract]:The Bayesian HAR latent factor model including time-varying coefficient and dynamic variance is constructed, and the high frequency realized volatility of Chinese financial futures (mainly stock index futures and treasury bonds futures) is forecasted. State potential factor model extraction includes volatility variables. The important information of the predictive variables, such as jump variables and symbolic jump variables considering leverage effect, is also included in the model. In order to investigate the influence of market speculation on volatility prediction in Chinese financial futures market, the prediction results show that the time-varying Bayesian latent factor model has the best short-term among all the forecasting models involved in the comparison. Medium and long term predicted effects. At the same time, Bayesian HAR family model with time-varying parameters and time-varying predictive variables greatly improves the prediction ability of fixed-parameter HAR family model. The speculative variables can be added to the forecasting model of stock index futures and treasury bond futures. Get better prediction results.
【作者单位】：华南理工大学工商管理学院;中山大学岭南学院;
【基金】：教育部人文社会科学研究青年基金资助项目(17YJC630099) 教育部人文社会科学研究规划基金资助项目(17YJA790011) 中央高校基本科研业务费广东省自然科学基金博士科研启动纵向协同管理试点资助项目(2017BQ014) 中国博士后基金面上资助项目(2017M612674) 广州市金融服务创新和风险管理基地资助项目
【分类号】：F724.5

【相似文献】