A reconstructive genetic algorithm in portfolio selection under concave transaction costs

In this paper we propose a hybrid algorithm allocated with genetic algorithm with branch and bound method for calculating a portfolio selection problem under concave transaction costs constraints. We study the multi-period investment and employ the absolute deviation of the rate of return of the portfolio as the measure of risk. In every period we solve linear programming subproblems by introducing linear underestimating function for concave transaction cost functions. Numerical experiments show that the algorithm solves this kind of problems efficiently.