Swarm intelligence-based stochastic programming model for dynamic asset allocation

Asset allocation is critical for the portfolio management process. In this paper, we solve a dynamic asset allocation problem through a multiperiod stochastic programming model. The objective is to maximise the expected utility of wealth at the end of the planning period. To improve the optimisation process of the model, we employ swarm intelligent optimisers, the Bacterial Foraging Optimisation (BFO) and the Particle Swarm Optimisation (PSO) algorithm. A hybrid optimiser using the BFO for initialisation and the Sequential Quadratic Programming (SQP) for searching the decision variables is also suggested. The results are compared with the stand-alone SQP and the canonical Genetic Algorithm. We have performed numerical experiments on 2-asset and 4-asset allocation problem respectively. The numerical results suggest that the hybrid method provides a better result especially for the 4-asset case, with improved fitness value and robustness than using BFO, PSO, GA, or SQP alone.