集成学习方法的已实现波动率预测和偏度信息含量研究  

Research on realized volatility forecasting and the information content of skewness for ensemble learning methods

作　　者：王云润 乔高秀[1] WANG Yunrun;QIAO Gaoxiu(School of Mathematics, Southwest Jiaotong University, Chengdu 611756, China)

机构地区：[1]西南交通大学数学学院,成都611756

出　　处：《重庆理工大学学报（自然科学）》2022年第4期243-253,共11页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(72001180);教育部人文社会科学研究项目(17YJC790119);中央高校基本科研业务费学科交叉研究专项(2682021ZTPY077);西南交通大学一流本科课程《金融数学》建设项目(YK20201140)。

摘　　要：为了探究风险中性偏度和历史偏度的信息含量及其对已实现波动率的预测能力,以S&P 500指数为研究对象构建出4个偏度指标,基于单个机器学习算法,提出数据驱动的窗口平均集成预测方法来克服市场结构突变导致的模型不确定性问题。经实证研究发现:风险中性偏度的预测能力整体上优于基于日数据和日内高频数据的历史偏度指标;在预测方法上,非线性支持向量回归(SVR)预测效果优于线性最小二乘估计和带惩罚项的线性回归,且基于SVR的窗口平均集成方法对波动率的预测效果最优。In order to explore the information content of risk-neutral skewness and historical skewness and their ability to predict the realized volatility,the article takes the S&P 500 index as the research object to construct four skewness indicators,and based on a single machine learning algorithm,a data-driven windows average ensemble forecasting method is proposed to overcome the model uncertainty problem caused by structural changes in the market.Empirical studies find that the predictive ability of risk-neutral skewness is generally better than historical skewness indicators based on daily data and intraday high-frequency data;in terms of forecasting methods,nonlinear support vector regression(SVR)has a better forecasting effect than the linear least square estimation and linear regression with penalty terms,and the windows average ensemble method based on SVR estimation obtains the best prediction ability.

关 键 词：已实现波动 偏度 S&P 500指数 机器学习 集成学习 

分 类 号：F830.9[经济管理—金融学]