Enhancing Financial Portfolio Robustness with an Objective Based on ϵ-Neighborhoods

Francisco Luna (1) , David Quintana (2) , Sandra García (3) , Pedro Isasi (2)

1 LCC - Departamento Lenguajes y Ciencias de la Computación
2 UC3M - Universidad Carlos III de Madrid [Madrid]
3 LADIS (CEA, LIST) - Laboratoire d'analyse des données et d'intelligence des systèmes (CEA, LIST)

Financial portfolio optimization is a challenging task. One of the major difficulties is managing the uncertainty arising from different aspects of the process. This paper suggests a solution based on ϵ-neighborhoods that, combined with a time-stamped resampling mechanism, increases the robustness of the solutions. The approach is tested on four of the most popular evolutionary multiobjective algorithms over a long period of time. This results in a significant enhancement in the reliability of the estimated efficient frontier.

Mots clés: GENETIC ALGORITHMS; MULTIOBJECTIVE EVOLUTIONARY ALGORITHM; Portfolio optimization; OPTIMIZATION; SELECTION; multiobjective optimization; robustness

Domaines: Gestion de portefeuilles [q-fin.PM]; Ingénierie assistée par ordinateur; Intelligence artificielle [cs.AI]; Système multi-agents [cs.MA]

Francisco Luna, David Quintana, Sandra García, Pedro Isasi. Enhancing Financial Portfolio Robustness with an Objective Based on ϵ-Neighborhoods. International Journal of Information Technology and Decision Making, 2016, 15 (3), pp.479-515. ⟨10.1142/S0219622016500115⟩. ⟨cea-01849801⟩