Topic
  • 3 replies
  • Latest Post - ‏2014-01-16T20:55:16Z by PaulRubin
mbaddar1
mbaddar1
2 Posts

Pinned topic MP application testing

‏2014-01-15T12:19:48Z |

Hello Forum members

i have an open question ( i think) we have a developed MP model and i am writing an application that uses the model to solve cost minimization optimzation problem , i am just wondering what are the typical methods for validating the results for such applications

 

thanks

  • PaulRubin
    PaulRubin
    58 Posts

    Re: MP application testing

    ‏2014-01-15T20:58:25Z  

     

    Are you asking how to verify the correctness of your model, or how to verify that the answers CPLEX is giving you are really optimal (and feasible)?

  • mbaddar1
    mbaddar1
    2 Posts

    Re: MP application testing

    ‏2014-01-16T07:43:06Z  
    • PaulRubin
    • ‏2014-01-15T20:58:25Z

     

    Are you asking how to verify the correctness of your model, or how to verify that the answers CPLEX is giving you are really optimal (and feasible)?

    Thanks Paul

    i am beginner in the area of MP so , if you can give me answers for both model correctness and test for results feasibility and optimality that would be of great help

    best regards

  • PaulRubin
    PaulRubin
    58 Posts

    Re: MP application testing

    ‏2014-01-16T20:55:16Z  
    • mbaddar1
    • ‏2014-01-16T07:43:06Z

    Thanks Paul

    i am beginner in the area of MP so , if you can give me answers for both model correctness and test for results feasibility and optimality that would be of great help

    best regards

    Typically what I will do is create one or more small instances of the problem, and set up a spreadsheet that takes potential solutions and computes their profit or cost and their impact on each constrained resource. In a network flow model, I will compute net flow into/out of each node in the spreadsheet, to verify flow balance. I then run the small instance(s) in CPLEX, put the solutions into the spreadsheet, and compare what the spreadsheet tells me is feasibility, objective value and resource utilization with what CPLEX tells me about those things. If there is a discrepancy, I need to figure out why the discrepancy arises, which hopefully will tell me whether the error is in the spreadsheet or the CPLEX model.

    This does not guarantee that my conceptual model for the problem (the "common denominator" of both the spreadsheet and the CPLEX model) is correct, but it helps me catch incorrect expression of the conceptual model.

    Paul