【主题】Multiple Forecasting based on Time Series PCA

【报告人】Qiwei Yao, 教授

英国伦敦政经大学

【时间】 2019年7月5日（星期五）14:00-15:00

【地点】上海财经大学统计与管理学院大楼1208会议室

【摘要】When forecasting large number of time series, the conventional wisdom is to forecast each time series separately, as the potential gain from looking into the cross correlations is typically cancelled out by the estimation errors.

We extend the principal component analysis (PCA) to vector time series in the sense that we seek for a contemporaneous linear transformation for a p-variate time series such that the transformed series is segmented into several lower-dimensional subseries, and those subseries are uncorrelated with each other both contemporaneously and serially. Therefore those lower-dimensional series can be forecasted separately. Technically it boils down to an eigenanalysis for a positive definite matrix. When the number of time series is large, an additional step is required to perform a permutation in terms of either maximum cross-correlations or FDR based on multiple tests. Numerical illustration with real data indicates that the forecasting based this new PCA outperforms those based on the original time series --- a phenomena can be clearly understood analytically.

The approach can be extended to forcasting multivariate volatility process, or a bundle of curve time series.

【嘉宾简介】姚琦伟，英国伦敦政治经济学院教授。姚琦伟教授1982年毕业于东南大学数学力学系，1987年于武汉大学获统计学博士学位。姚琦伟教授的主要研究领域为：非线性及线性时间序列分析、非参数回归分析、时空过程分析、经验似然及bootstrap方法,以及统计在金融、经济等方面中的应用。

【主持人】尤进红