Enhanced multi-objective evolutionary algorithms for large-scale portfolio optimization

The paper presents strengthening strategies for multi-objective evolutionary algorithms to address large-scale portfolio optimization under cardinality constraints, where exact methods become inefficient due to NP-hard complexity.

• Introduces a unique solution representation, a novel operator, and new repair mechanisms to handle lower and upper limits on the number of assets.
• Implements customized mating strategies within well-known multi-objective evolutionary algorithms.
• Tests the proposed approach against traditional algorithms using well-known market indices as benchmarks.

Results indicate that the proposed strategy provides better approximations and converges faster without loss of performance as the number of assets increases.