Portfolio Selection Problems

Abstract

the portfolio management is an interesting problem Treated, modeled and solved by several expert over the world. In this work, we combine recurve neuronal network tool and an evolutionist method to solve this problem under his multi-objective form: minimizing the portfolio risk and maximizing the portfolio return, with a cardinality constraint. To solve the proposed model, the continuous Hopfield neural network is used, basing on an original energy function, starting from several initial solutions; to improve the decision quality, the obtained solutions set is presented to the multi-objective genetic algorithm as an initial population. At last, we compare our solution to the genetic algorithm and continuous Hopfield network solutions. The proposed strategy has proven his robustness in the experimental results section.