Font Size:
Cardinality Constrained Portfolio Optimization by Means of Genetic Algorithms
Last modified: 2009-08-30
Abstract
When applying the standard
Markowitz mean-variance model on a real
portfolio selection problem, we are faced with
certain limitations like cardinality of chosen
assets, discrete nature of trading variables etc.
While the classical mean-variance model can be
successfully solved by standard algorithms
(quadratic programming), modelling an actual
investment leads to NP-hard optimization
problem. In such circumstances heuristic
methods appear as the only way out.
This paper aims at finding efficient
evolutionary inspired algorithm for cardinality
constrained portfolio optimization. Among
developed algorithms which were able to solve
problems with very many possible assets, the
algorithm with hybrid crossover presents itself as
the most effective. In order to make obtained
results comparable, test sample was chosen from
databases that serve as a benchmark for this
problem class.
Markowitz mean-variance model on a real
portfolio selection problem, we are faced with
certain limitations like cardinality of chosen
assets, discrete nature of trading variables etc.
While the classical mean-variance model can be
successfully solved by standard algorithms
(quadratic programming), modelling an actual
investment leads to NP-hard optimization
problem. In such circumstances heuristic
methods appear as the only way out.
This paper aims at finding efficient
evolutionary inspired algorithm for cardinality
constrained portfolio optimization. Among
developed algorithms which were able to solve
problems with very many possible assets, the
algorithm with hybrid crossover presents itself as
the most effective. In order to make obtained
results comparable, test sample was chosen from
databases that serve as a benchmark for this
problem class.
Full Text:
PDF