The Application of Robust Optimization and Goal Programming in Multi Period Portfolio Selection Problem

Portfolio selection is one of the most important area in financial world. Investors always want to make the best decisions which are compatible with conditions of real world. In the real world, data are usually under uncertainty. On the other hand, the most of strategies for portfolio selection are multi-period. Therefore, investors should rebalance their portfolios during investment horizon. In this research we present a multi-period portfolio optimization model which considers transaction costs and deal with uncertainty by application of robust programming. This model is a mean-CVaR multi objective model that is solved by goal programming. Furthermore, most of previous researches have used regression or time series models to forecast future returns of stocks for solving numerical examples, however, in this paper we forecast future returns by using Artificial Neural Networks (ANNs). Finally, solutions of robust model are compared with results of nominal one. These results show that consideration of data uncertainty and other real assumptions lead to more practical solutions.