Weighted Estimation Method of High Dimension Portfolio Allocation for Time-Varying Stock Market

发布时间：2025-03-17 作者：浏览次数：

Speaker:	黄磊	DateTime:	2025年3月22日（周六）上午11:00-12:00
Brief Introduction to Speaker: 黄磊，2015年博士毕业于新加坡国立大学，现任职于西南交通大学数学学院统计系，副教授，研究生导师，科研兴趣包括半参数回归模型、时间序列分析、生物统计。在国内外学术期刊The Annals of Statistics, Journal of Business & Economic Statistics, Energy, Statistics in Medicine, Journal of Nonparametric Statistics, Computational Statistics & Data Analysis等发表论文20余篇。主持国家自然科学基金青年项目1项、省级项目3项、校级项目6项。2020年获四川省数学会首届应用数学奖一等奖。2021年入选中国现场统计研究会理事。		Place:	国交2号楼315会议室
Abstract:The primary goal of portfolio allocation is to achieve higher returns while controlling risk. In the background of large portfolios, the classical Mean-Variance Portfolio (MVP) model faces high-dimensional challenges, which severely reduces the accuracy of the estimation. Meanwhile, policy changes and market structure shifts may contribute to time-varying stock market to some extent. To address the poor performance of existing methods based on the classical MVP model in the time-varying stock market, this article proposes a Time-Varying Weighted Large Portfolio (TVWLP) method. It is grounded in the factor model and introduces various weight functions to adjust the estimated factor loading matrix, and timely adjust the mean vector and covariance matrix. Furthermore, we establish the asymptotic properties of the timely adjusted method under some regular conditions. We validate the effectiveness of TVWLP through some simulations. Finally, in empirical studies, we demonstrate the superi...

上一条： Singularity type of the long-time Kahler-Ricci flow
下一条： Community detection in weighted networks via the profile-pseudo likelihood method