GA-MEPS: Multiple Experts Portfolio Selection Based on Genetic Algorithm

In the realm of online portfolio selection, aggregating multiple experts is crucial for improving investment decisions in complex market conditions. However, existing methods often overlook two essential factors: expert diversity and adaptive expert selection, both of which significantly impact portfolio returns. To address these issues, this paper proposes a novel multiple experts portfolio selection method based on genetic algorithm (GA-MEPS). This method (i) introduces a portfolio trend tracking problem with closed-form sparse solutions, enabling optimal selection among assets within a unified framework; (ii) constructs a pool of 24 experts, each characterized by distinct trading behaviors; and (iii) designs a fitness function to enable efficient adaptive expert selection based on genetic algorithm. Extensive experiments on six datasets demonstrate GA-MEPS’s superior performance over six competitive algorithms, achieving average cumulative wealth and Calmar ratio that are 4.17E+09 and 5.55 times higher, respectively, compared to the benchmark buy-and-hold strategy.