ROBUST MEAN-VARIANCE PORTFOLIO SELECTION WITH WARD AND COMPLETE LINKAGE CLUSTERING ALGORITHM

In this paper, we present a robust mean-variance portfolio selection method with preprocessing data using cluster analysis. Using this proposed method, we obtain the best portfolio (i.e. portfolio with the highest Sharpe ratio) efficiently when there is a large number of stocks involved in the formulation of the portfolio. On the other hand, this procedure is also robust against the possibility of outliers existence in the data. Based on our empirical study, we find that the performance of portfolio produced using clustering with Ward algorithm is better than portfolio performance produced by the clustering with complete linkage algorithm for all risk aversion values gamma. Besides, we also find that portfolio performance with robust FMCD estimation is better than portfolio performance with robust S estimation and classic MV portfolio for all risk aversion values gamma, for both portfolios produced by cluster analysis with Ward algorithm and complete linkage algorithm.