Parallel MOEAs for Combinatorial Multiobjective Optimization Model of Financial Portfolio Selection

This paper proposes a multiobjective optimization model with rebalancing and cardinality constraint to obtain portfolios consistent with practical aspects of the financial investment process. To perform this optimization, parallel versions of the evolutionary algorithms PDEA, SPEA2 and NSGA-II are proposed, aiming a low execution time in accordance with the rapid oscillations in the stock market. In-sample analysis compares sequential and parallel versions of evolutionary algorithms while out-of-sample analysis performs simulations using different algorithms and models, seeking to identify the effect of the rebalancing process for the model. Results point out the importance of the model with rebalancing and the efficiency of the parallel algorithms, mainly in relation to the execution time.