Portfolio cardinality constraint in a genetic algorithm

Hello,

does anybody have a good way to include a portfolio cardinality constraint in an elegant way in a genetic algorithm? I want to form a portfolio of exactly 50 stocks out of a set of 300 stocks, each of those in the portfolio should have a weighting between 1% and 3% and the others need to be forced to zero.

Happy to see your ideas :-)

Answer 1

[Alan Weiss]

Well, not in a genetic algorithm, but you might find this example using MILP to approximate MIQP to be relevant.

Alan Weiss

MATLAB mathematical toolbox documentation

[Christian Schwehm]

Hi Alan,

cheers for the hint, but I cannot really use a MILP, as I want to run the heuristic with a rather complex fitness function. I could include a penalty term for any deviations from the "correct" cardinality, but the GA would need to do a whole lot of "useless" computations, that could be saved, if it were possible to restrict the number of weights>0 to a fixed number...

Any ideas?

Christian Schwehm