Optimization of cardinality constrained portfolios with a hybrid local search algorithm

One of the main advantages of portfolios over single assets is that risk can be diversified without necessarily reducing the expected return – provided “proper” assets are selected and they are assigned the “proper” weights. Since in practice investors tend to restrict themselves to a rather small number of different assets, the decision which securities to include is a crucial one that turns out to be NP-hard. In this paper we suggest a hybrid local search algorithm which combines principles of Simulated Annealing and evolutionary strategies and which proves to be a highly efficiently approach this problem.