Improved Portfolio Optimization Combining Multiobjective Evolutionary Computing Algorithm and Prediction Strategy

In conventional mean-variance model of portfolio optimization problem the expected return is taken as the mean of the past returns. This assumption is not correct and hence the method leads to poor portfolio optimization performance. Hence an alternative but efficient method is proposed in which the mean and variance of expected return are first predicted with a low complexity functional link artificial neural network model (FLANN). The predicted values of mean and variance are consequently used in multi objective swarm intelligence techniques for achieving better performance. The multi objective swarm intelligence techniques chosen are non-dominated sorting genetic algorithm-II (NSGA – II) and multi objective particle swarm optimization (MOPSO).The performance of the proposed prediction based portfolio optimization model has been compared with the Markowitz mean-variance model. The comparison of the performance includes three performance metrics, Pareto front and nonparametric statistical test using the Sign test. On examining the performance metrics it is observed that the proposed prediction based portfolio optimization model approach provided improved Pareto solutions but maintaining adequate diversity.