Prev: Merge on multiple variables
Next: Analytics
From: ankur on 4 May 2010 09:30
Hi,

I need to obtain top 100 solutions to a LP problem where all the
decision variables are binary. For this I have written a code that
works as follows -

1. Solve the LP (maximize SOME OBJECTIVE subject to CONSTRAINT1
CONTRAINT2 etc.). This gives me the best solution to the problem.

2. Add another constraint to above LP which excludes the solution
obtained in previous step (to obtain the next best solution) form
feasible region.

3. I repeat above process to obtain top 100 solutions

I am using following code snippet for solving the LP -

proc lp data = lp_definition sparsedata imaxit=10000
primalout=lp_primal_&i.;
run;

I see that in some of the iterations, I do not get an optimal
solution. Re-running the code after sometime gives me an optimal
solution.
I think this is happening because the procedure terminates before
finding an optimal solution. I tried to solve that by adding
"imaxit=10000" to PROC LP statement, but this does not help.

Can someone please help me in specifying any option that guarantees a
successful solution (run time is not a constraint).

Thanks,
Ankur Bohra.
 | 
Pages: 1
Prev: Merge on multiple variables
Next: Analytics