Ant Colony Optimization in Multiobjective Portfolio Selection

Autor(en)

Karl Franz Dörner, Walter Gutjahr, Richard Hartl, Christine Strauss, Christian Stummer

Abstrakt

Multiobjective decision-making and combinatorial optimization have been studied extensively over the past few decades. Both fields play a decisive role in multiobjective combinatorial optimization, for which the class of (multiobjective) portfolio selection is of particularly high practical relevance. Support for making multiple objectives decisions is based on either (i) approaches aggregating different types of benefits (e.g., cash flow, sales or even intangibles such as image) in a unique overall objective function or (ii) approaches that (partially) determine the efficient (i. e., Pareto-optimal) portfolio candidates and then allow the decision-maker to interactively "move" in that solution space until a satisfactory alternative is found. While the former techniques often require extensive a priori preference information, those of the latter school regularly can do without such data. However, the process involved in identifying efficient portfolios is not trivial. While a brute-force complete enumeration procedure can determine them within acceptable time for comparatively small problems only, that task becomes increasingly demanding as the number of projects grows. In such complex problems, heuristic-based approaches can provide an attractive trade-off between the quality of the approximation of a solution space and the computing capacity required to achieve this approximation. Our approach applies a constructive meta-heuristic, Ant Colony Optimization (ACO), that imitates the behavior shown by real ants when searching for food. In order to meet multiobjective problem specific requirements the adapted ACO approach implements several pheromone matrices and random weights for their use. The lifespan concept and the pheromone decoding scheme are two more novel features which are necessary to model the portfolio selection process.

Organisation(en)

Institut für Business Decisions and Analytics, Institut für Rechnungswesen, Innovation und Strategie, Institut für Soziologie, Forschungsgruppe Data Mining and Machine Learning, Institut für Statistik und Operations Research

Seiten

243-248

Publikationsdatum

2001

Peer-reviewed

Ja

ÖFOS 2012

502052 Betriebswirtschaftslehre, 101015 Operations Research

Link zum Portal

https://ucrisportal.univie.ac.at/de/publications/bc542e4e-debc-4873-876c-3e4d7f1ef1f1