Originally posted by: Nguyenducduy

Dear all,

I have a VRP problem and i tried to solve it in 2 ways:

Default Behaviour

Setting an initial solution

But the running time is not significantly difference between 2 methods. 
 when i saw engine log of both methods. 70% running time, Cplex tries to find the solution when MIP gap <5%. So that, Setting an initial solution method only help improve running time from MIP gap 100% to 5%.

Do you have any idea to improve my problem? Thank you very much. 

An initial solution will not cause the lower bound to improve faster. The best way to close the gap faster may be to find a tighter formulation (easier said than done, I know) or to find helpful user cuts that you can add with a callback (again, easier said than done). Another possibility would be to stop the solver at some point (say, when the gap drops below 5%) and then switch the MIP emphasis parameter setting to either proving optimality or tightening the best bound. Then call the solve command again. CPLEX will resume where it left off, but with the new emphasis. (This is not guaranteed to help, but it's something I try from time to time.)

Originally posted by: Nguyenducduy

Thank you very much. I will try to do it.