Search code examples
mathematical-optimizationsolverlinear-programmingms-solver-foundation

Solver Foundation in C# has a bigger error than Excel's solver

I'm starting to use Solver Foundation in a WPF/C# app, which should replace an Excel sheet that solves a linear problem, something simple like this:

  • Mix A: (20% A + 70% B + 10% C)
  • Mix B: (35% A + 65% C)
  • Mix C: (10% A + 80% B + 10% D)

How much do I need of each mix to get as closest as possible to (15% A + 70% B + 10% C + 5% D).

Pretty simple, even for Excel. So... I create this model in an OML string, and solve it with Solver Foundation, but the results are not the same as I get in Excel, and in every case the quadratic error I get is bigger with the Solver Foundation results (checked in the Excel sheet).

Is there any way I can configure the solver to get the same result as in Excel? If you need to see the OML, please ask and I'll update the question.

Solution

  • Are you sure you are attempting to minimize the same result?

    Maybe the two methods are using a different difference measurement.

    For instance you seem to be measuring R^2 as your solution, is that what your C# code is using as a measure of distance from perfect?