As my question states, the problem I am having is finding a sensible way to search a large space. Any help or insight that could be provided would be hugely appreciated.

Currently I am trying to search through a space of possible investment strategies. This space has been restricted to 3 possible assets (Equity, Cash and Bonds) across 100 years where strategies are constant for 10 years at a time. I have also constrained the area by only allowing 100% investment in one asset or a 50/50 split between two. All of the fund must be invested at each time. This means that there are 10 times to choose between 6 possible combinations of investments. I have already preformed a basic search of the space using a labour intensive method of Genetic Algorithms which requires me to choose suitably optimal strategies across larger times, working my way down to decades starting at 50 year blocks. However I believe that there would be a better solution but my lack of knowledge in optimization is hindering me here. A strategy is ran through a programme to provide quantiles of a fund life which are then matched to given criteria.This is what indicates the suitability of a strategy.

I have been looking into using Bellman's equation but since it requires markovian assumptions I don't know if I can apply this. If anyone has any ideas that could be applied it would be a great help.

If anything I've said requires clarification or if I have been a bit vague on some information please let me know.

Thanks in advance.

  • 1
    $\begingroup$ I believe the best way is really to use bellman equations. You need to specify an utility function, then you can approximate returns on each asset using gaussian quadrature methods and use a search grid for the choice variables (i.e. how much to invest on each asset). If you have a terminal year, you can easily solve it by backward induction. $\endgroup$
    – phdstudent
    Jul 20, 2015 at 11:41


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.