Publication at Faculty of Mathematics and Physics |
2010
Abstract
The portfolio selection problem using mean-risk models is formulated for several risk measures (variance, VaR, CVaR, absolute deviation and semivariance) and for normal, Student and discrete distributions of returns. An extensive numerical study investigates the convergence properties of solutions obtained for sample-based approximations to solutions for continuous distributions and its sensitivity to the chosen risk measure and the initial probability distribution.
Cluster analysis is applied to selection of the best approximate solution. The computational part of this work is realized in C++ language, while using GAMS for optimization.