Improved portfolio optimization with non-convex and non-concave cost using genetic algorithms

In this paper, genetic algorithm is proposed to solve an improved portfolio optimization model effectively whose objective function is both non-convex and non-concave. The practical constraints, such as minimum trading unit, transaction cost and money limits, have been taken into consideration in this improved portfolio optimization model based on the traditional Markowitz model. The introduction of non-convex and nonconcave typical transaction costs makes the model more representative. This problem is a typical NP-hard problem because of the introduction of non-convex and non-concave objective function which can better meet the practical investment conditions. The empirical results of Chinese stock market have shown both the applicability of the improved model and the effectiveness of GA.