Originally posted by: Ulysses.Sheng

Hello,

Could anyone tell me how to use GetObjSa in the cplex concert technology for C++?
suppose it is: min c1*x1+c2*x2
s.t. Ax<=b
I am wondering how to write the codes exactly to get the analysis for c1 and c2.

Thank you all in advance!
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

Here's a code segment that reads in an MPS, LP or SAV file, extracts the
model, solves the model, then performs objective sensitivity analysis
using IloCplex::getObjSA. In the first two arguments, you pass in the arrays
that will contain the lower and upper limits of the objective coefficients for
which you want sensitivity analysis. The third argument is the array of
IloNumVars for which you want objective sensitivity analysis.
IloObjective obj;
IloNumVarArray var(env);
IloRangeArray rng(env);
cplex.importModel(model, name, obj, var, rng);

cplex.extract(model);
if ( !cplex.solve() ) {
env.error() << "Failed to optimize LP" << endl;
throw(-1);
}

cplex.out() << "Solution status = " << cplex.getStatus() << endl;
cplex.out() << "Solution method = " << cplex.getAlgorithm() << endl;

IloNumArray objLB(env);
IloNumArray objUB(env);
cplex.getObjSA(objLB, objUB, var);
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Ulysses.Sheng

Hey, dear EdKlotz,

I have another question to ask you.
If I have a base case value for all the coefficients of objective function, and I want to know by varying within what limit, the optimal decision doesn't change. How should I change the codes?

i mean if i know c1 and c2 and get an optimal decision in terms of x. Then i want to know within what range of change of c1 or c2, the optimal decision is the same.

Thank you!

Ulysses
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: EdKlotz

> Ulysses.Sheng wrote:
> Hey, dear EdKlotz,
>
> I have another question to ask you.
> If I have a base case value for all the coefficients of objective function, and I want to know by varying within what limit, the optimal decision doesn't change. How should I change the codes?
>
> i mean if i know c1 and c2 and get an optimal decision in terms of x. Then i want to know within what range of change of c1 or c2, the optimal decision is the same.
>
> Thank you!
>
> Ulysses

getObjsa gives you the range of change of individual elements of the objective.
In other words, it tells you how much you can change a single element of the objective without losing optimality (i.e. dual feasibility). You can also view
this as performing parametric analysis on the objective with a unit vector.
In other words, the element in the unit vector with a 1.0 coefficient identifies
the individual objective coefficients that for which you want to assess the
maximum change without compromising optimality.

So, are you now asking how to calculate the amount you can change multiple
objectives at once (i.e. parametric analysis on the objective for an arbitrary
vector rather than a unit vector)?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Ulysses.Sheng

Hey, dear EdKlotz,

My question is: suppose the problem
min c1*x1+c2*x2
s.t A*x<=b

and in the base case c1=100

but when I use:
IloNumArray objLB(env);
IloNumArray objUB(env);
cplex.getObjSA(objLB, objUB, x);

it gives me objLB[0]>100

Because I thought it should be like lower bound<100 and upper bound>100?

thanks!
Ulysses
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Ulysses.Sheng

Hey, dear EdKlotz,
sorry I made a silly mistake there.
It gave the right answer actually.
Thanks for you great help!!
Ulysses
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: Ulysses.Sheng

Hey dear EdKlotz, thanks:)
it works!
#CPLEXOptimizers
#DecisionOptimization