MULTIOBJECTIVE DYNAMIC OPTIMIZATION OF INVESTMENT PORTFOLIO BASED ON MODEL PREDICTIVE CONTROL

In this paper, a multiobjective model predictive control (MO-MPC) for portfolio selection is proposed. The objective functions are defined using a multiperiod format through the receding horizon strategy, considering the expected wealth, the variance, and the conditional value at risk as the objective function to be optimized, including transaction costs, self-financing, and investment limits for each asset. A Pareto front is obtained in each step by a multiobjective genetic algorithm, and a Pareto optimal point is chosen as the control action applied to the system. This choice is made based on a selection criterion according to the investor profile. Finally, the performance of the MO-MPC facing extreme situations of the financial market is investigated through numerical experiments using data from the Brazilian stock exchange. Simulation results show that the MOMPC can deal with the dynamic and unstable scenario, efficiently tracing the trade-off between risk and return, managing the transaction costs and bounds.