机构地区: 温州大学数学与信息科学学院
出 处: 《统计与信息论坛》 2011年第6期18-22,共5页
摘 要: 将lasso图理论合并到状态空间模型中,利用条件独立性且通过范数惩罚法对协方差阵进行估计。新方法兼具图模型和动态状态空间模型的优点。最后将该方法应用于欧洲股票市场进行投资组合优化决策,结果表明基于lasso图方法的状态空间模型的投资组合业绩要优于自回归和一般的状态空间模型。 The state space model based on graphical methods with LASSO is introduced and the covariance matrix of the residual is estimated.The proposed method combines the advantages of undirected graphical models and dynamic state-space models.The application to European stock market shows that our models generate higher profits in the long run comparatively to the Vector Autoregressive model and ordinary state space models.