國立政治大學統計學系
學 術 演 講
學 術 演 講
主講人：黃士峰 教授(中央大學統計研究所)
題 目：Portfolio Optimization via Dynamic Networks and Vine Copulas
時 間：民國114年5月12日 (星期一) 下午1：30
地 點：國立政治大學逸仙樓050101教室
摘 要：
This study explores the application of vine copulas combined with network-based methods for portfolio optimization. A de-GARCH technique is employed to preprocess each series to address inherent characteristics such as autocorrelation, conditional heteroscedasticity, and volatility clustering in financial time series. A similarity matrix is then computed from the multivariate de-GARCH data and used to construct a global minimum spanning tree (MST), which facilitates the identification of suitable stocks for portfolio construction. A local MST (LMST) is subsequently built from the selected stocks, and various vine copulas are applied based on the LMST structure to model their joint distribution. This copula-network-based distribution is then used to determine the portfolio weights. The empirical analysis, conducted on component stocks of the S&P 100 index over the 2019–2023 period using a rolling-window framework, shows that the proposed method achieves competitive cumulative returns compared to benchmark approaches.
