I would be biased towards using a heuristic, for instance Threshold
Accepting (TA), for solving such a problem. (TA is implemented in
package NMOF. Disclosure: I am the author of that package.) But you will
not find a ready-to-use solution there.
(1) you need an objective function, ie, a function that maps a given
vector of holdings (and data like your scenario matrix) into a real
number; the better the portfolio, the lower the number.
(2) For TA, you need a so-called neighbourhood function. That is a
function that changes one portfolio vector into another, by changing
some elements. Examples for simple neighbourhoods are in the package
vignettes. Do you have a budget constraint? If yes, and you want to work
with integers, I would suggest using a "cash" variable. (See, eg,
Algorithm 3 in http://www.swissfinanceinstitute.ch/rp20.pdf )
(3) The constraints can, at least in a first round, be included through
penalties.
Regards,
Enrico
PS. There is a mailing list dedicated to finance-with-R questions, and
you may get better answers there.
https://stat.ethz.ch/mailman/listinfo/r-sig-finance
--
Enrico Schumann
Lucerne, Switzerland
http://nmof.net/
Am 13.01.2012 17:06, schrieb Sal Pellettieri:> Hi,
>
> I'm an R newbie and I've been struggling with a optimization
problem for
> the past couple of days now.
>
> Here's the problem - I have a matrix of expected payouts from different
> stock option strategies. Each column in my matrix represents a different
> stock and each row represents the return to the strategy given a certain
> market move. So the rows are not a time series of percentage returns but a
> dollar payout in different expected scenarios, i.e.
>
> Expected Return Matrix (ER) = stock1 stock2 ....
stockn
> scenario1 $
$
> $
> scenario2 $
$
> $
> scenario3 $
$
> $
> ...
>
> I want to create an optimal portfolio of these strategies by applying a
> vector of weights. The weights will be the number of contracts of each to
> buy and won't be a percentage weighting. There are a few constraints I
need
> it comply with:
>
> - The weights have to be integers
> - The minimum portfolio return (ER* Weights) across the scenarios has
to
> be greater than some negative number I specify
> - There has to be a certain minimum number of stocks in the portfolio
so
> length(weights)>some number I specify.
>
> Any help is GREATLY appreciated since I have tried so many different
> functions and packages. Even if someone can just lead me to the correct
> function to use that would be a great help as I've looked at optim,
> solveLP, ROI package and many others.
>
>
> Thanks,
> S
>