A Novel Constraint Handling Technique for Complex Portfolio Selection

Most researches on portfolio selection problems only consider simple boundary constraints, which can be solved by normalized method. In some cases, investors may purchase a stock under complex boundary constraints. At this situation, the normalized method is invalid. In this paper, a new approach is proposed to deal with complex boundary constraints. The main idea of the new approach is to modify the normalized method and transfer violated components to other feasible components. The proposed method can be applied to many algorithms, and this paper only presents an application on particle swarm optimization (PSO). Experimental results on a realistic portfolio selection problem in China show that the proposed method can not only successfully handle the complex boundary constraints, but also achieve good performance.