Optimal financial portfolio selection using a metaheuristic approach with multiple strategies

Abstract

Portfolio optimisation with cardinality constraints has been extensively studied in the realm of financial investment, recognised as an NP-hard quadratic programming problem. As an innovative metaheuristic approach, the dung beetle optimiser leverages its unique optimisation search mechanism to effectively tackle unconstrained optimisation problems. However, the realities of portfolio optimisation involve various constraints; thus, the original dung beetle optimiser may not suffice. Consequently, this study develops an improved dung beetle optimiser to address cardinality constrained portfolio optimisation, incorporating a new decision variable update strategy, a constraint handling strategy, and a local search strategy. These techniques facilitate the efficient selection of assets from among multiple candidate assets. To validate the capabilities of the indicated methodologies, five datasets from OR-Library and six datasets from NGINX are employed for testing. The results from these datasets consistently indicate that the proposed strategies outperform existing alternatives. Furthermore, the comparison results with various methods presented in other works demonstrate that the proposed technology is competitive in the realm of cardinality constrained portfolio optimisation.