Decision Optimization

Decision Optimization

Delivers prescriptive analytics capabilities and decision intelligence to improve decision-making.


#Analytics
#DecisionOptimization
#DecisionOptimization
 View Only
  • 1.  MIP problem not reaching global optimum

    Posted 01/19/16 08:39 AM

    Originally posted by: MooseGreen


    Hi,

    I have an MIP model in OPL that is quite simple. However, I noticed the solution sub-optimal, depending on how I set one of the parameters in the model.

    In short, I am using parameter TranCost to suppress the number of transactions, with the help of binary variable BinSDeliver.

    Below are my test scenarios:

    Scenario 1: TranCost = 0. The result is optimal, with the number of transaction equal 4.
    Scenario 2: TranCost = 0.5. The result is optimal, but with fewer transactions (3 to be exact). The objective function (excluding the transaction cost) is 11953.4035833763
    Scenario 3: TranCost = 1. The number of transactions is still 3, but it is now with a suboptimal result. The objective function (excluding the transaction cost) is 11953.6701255976. At least one better solution is available, where the allocation is the same as when TranCost = 0.5.

    Why would CPLEX not find the global minimum for such a simple problem, when TranCost = 1?

    Have attached the model and data file. Would appreciate any input.

     

    Moose


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 2.  Re: MIP problem not reaching global optimum

    Posted 01/19/16 01:48 PM

    Hi,

    this looks like tolerance matter.

    Can you try with

    execute PARAMS {
      cplex.epgap = 1e-8;
     
    }

    regards


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 3.  Re: MIP problem not reaching global optimum

    Posted 01/20/16 03:11 AM

    Originally posted by: MooseGreen


    Hi Alex,

     

    Thanks for the suggestion. I tried that, but it gave the exact same results as 1e-5. I have also tried as small a value as 1e-19, but it didn't change anything. 

     

    I have also tried to set repeatpresolve = 0, but it also didn't change anything.


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 4.  Re: MIP problem not reaching global optimum

    Posted 01/20/16 03:44 AM

    Hi

    have you replaced or added the new setting ?

    Which version do you use ?

    regards


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 5.  Re: MIP problem not reaching global optimum

    Posted 01/20/16 03:48 AM

    Originally posted by: MooseGreen


    Hi,

    I use 12.6.0.0.

     

    I have added the below to my .mod.

    execute PARAMS {
      cplex.epagap = 1e-19;
      cplex.repeatpresolve = 0;
    }


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 6.  Re: MIP problem not reaching global optimum

    Posted 01/20/16 03:58 AM

    Hi,

    I suggested epgap. But you seem to use epagap.

    CPLEX 12.6.3 is the latest version

    regards


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 7.  Re: MIP problem not reaching global optimum

    Posted 01/20/16 04:03 AM

    Originally posted by: MooseGreen


    Hi,

    That works! Sorry I mistook epagap for epgap...

    Many thanks for your help, Alex!


    #DecisionOptimization
    #OPLusingCPLEXOptimizer


  • 8.  Re: MIP problem not reaching global optimum