EcoSta 2019: Start Registration
View Submission - EcoSta2019
Title: Large portfolio management with clustering techniques Authors:  Huei-Wen Teng - National Chiao Tung University (Taiwan) [presenting]
Abstract: Large portfolio management faces many numerical problems and statistical difficulties. For instance, it is non-trivial to estimate a large covariance matrix that remains semi-positive, it is time demanding in the optimizing process when the dimension is high, and the optimized portfolio may not be stable or with high turnover rate. Instead of proposing an alternative approach to estimate the large covariance matrix, the aim is to propose a clustering method to overcome the above problems. With hierarchical clustering techniques, we partition the assets into several groups, so that assets behave similarly within groups but vary among groups. In each group, the asset closest to its centroid is selected as the candidate asset. The optimization procedure is then implemented for the selected small portfolio. With empirical analysis, we will show that the proposed method is comparable with that optimized directly from the large portfolio.