Originally posted by: SystemAdmin

[cwinter said:]

Hello. I am running a timetabling problem of OPL version 5.5.1. I have 55000 constraints and i need to know the shadow prices of each one in order to make changes, and get to a better solution. How do i do this?. I´ve tried changing the decision variable from int in 0..1 to float+ and naming the contraints like forall(......)
CT:
Constraint <=W for example but it only shows "open values", "open slacks".<br />
I am very new at this, even though the programming is great.

Thank you very much.
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[alain.chabrier said:]

Hi,

are you sure you are not including in your model something that makes it a MIP (integer variable or  logical constraint) ?

I just tried with :
range r = 1..10;
dvar float+ x[r];

minimize sum (i in r) i*x[i];
subject to  {
  forall(i in r)
  ct:x[i] >= i;
  }

and I get the duals as a choice in the ocntextual menu of the constraints in the problem browser.


Alain
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[cwinter said:]

Alain, thank you. I´ve seen examples and i´m sure its MIP. I will attach the code at night because i don´t have it with me right know.

Is the any code i can write after the execute {  command that would show me the shadow costs or dual values of the constraints on the console?

thank you
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[alain.chabrier said:]

If it is a MIP you will not have dual values, it needs to be an LP.

Alain
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[cwinter said:]

It is LP. Here is the code.

{string} Cursos = ...; //total de cursos a programar
{string} Dias = ...; //total de dias
int cruces[Cursos][Cursos] =...; //numero de alumnos que toman par de cursos
//int cruces1[Cursos][Cursos] =...;
int creditos[Cursos][Cursos]=...;
range Semestres = 1..12;
int CurT[Cursos][Semestres] = ...;//1 si el curso pertenece al semestre

tuple structCursosCursos {string c1; string c2;};
//tuple cursoDiaType {string Cursos; string Dias;};
float duals[Cursos] ;
int SiVariable[Cursos][Cursos] = ...;//parametro auxiliar para contruir las tuplas cursos cursos para las cuales existira la variable y que indica si tienen examen el mismo dia
int SiDia[Cursos][Dias] = ...; //binaria que indica si curso puede ser dictado en un dia por disponibilidad de profesor
float horashoras[Cursos][Dias]=...;
//int SiSa2[Cursos] = ...; //binario si profesor puede tomar examen sabado2

{structCursosCursos} CursosCursos = {<c1,c2> | c1,c2 in Cursos: SiVariable[c1][c2] == 1};

{structCursosCursos} CursosCursosSemestre = {<c1,c2> | c1 in Cursos,c2 in Cursos, t in Semestres: (CurT[c1][t]*CurT[c2][t] == 1) && ord(Cursos, c1) < ord(Cursos,c2) };<br />
{structCursosCursos} CursosCursosSemestreB[t in Semestres] = {<c1,c2> | c1 in Cursos,c2 in Cursos: (CurT[c1][t]*CurT[c2][t] == 1) && ord(Cursos, c1) < ord(Cursos,c2) };<br />
dvar float x[Cursos][Dias] in 0..1; //si curso se dicta en dia
//dvar boolean y[CursosCursos]; //si pares de cursos tienen examen el mismo dia
dvar boolean l[CursosCursos]; //si dos pares de cursos tiene examen en dias consecutivos
dvar float+ g[CursosCursos]; //nuemro de horas diff entre par de cursos
dvar float+ h; //total de horas entre cursos para un solucion especifica
dvar float+ horastotales;
dvar int d[Semestres]; //Cuenta de Alumnos de un mismo semestre con exámenes en días sucesivos
dvar int m; //Cuenta de Alumnos con exámenes en días sucesivos
dvar int c;
dvar int horas[CursosCursos];
dvar boolean w[CursosCursosSemestre]; //si dos pares de curos de un mismo semestre tiene examen en dias consecutivos
dvar int a[Semestres] in 0..1; //una posibilidad de cursos de un mismo semestre consecutivos muy penalisada
//{cursoDiaType} CursosDiasCombinations = {<i,d> | t in Semestres, d in Dias, i in Cursos: x[i][d] > 0 && CurT[i][t] == 1};



minimize ( m //+c/64
+sum (t in Semestres) a[t] 
//+sum (t in Semestres) d[t]
//+h
);

subject to{

  //todos los cursoso tienen un solo examen
  forall (i in Cursos)
  CT:
    sum(d in Dias) x[i][d] == 1;
 
  //Cursos que no se pueden dar un sabdo
  sum(i in Cursos, d in Dias) x[i][d]*(1-SiDia[i][d]) == 0;

  //Cursos con cruces no pueden tener examen un mismo dia
  forall (i in CursosCursos, d in Dias: cruces[i.c1][i.c2]> 0)
    CT2:
        x[i.c1][d] + x[i.c2][d] <= 1;<br />        //CT2[x].dual;
   
    //No mas de una pareja de cursos de un mismo semestre con Pruebas en dias consecutivos
  forall (i in CursosCursosSemestre, d in Dias: ord(Dias, d) < 19)<br />CT3:       
x[i.c1][d] + x[i.c1][next(Dias, d)] + x[i.c2][d]+ x[i.c2][next(Dias, d)]<=1 + w[i];  <br />  forall (t in Semestres)
CT4:   
    sum (i in CursosCursosSemestreB[t]: cruces[i.c1][i.c2]> 0) w[i]<= a[t];<br />  forall (t in Semestres)
        sum (i in CursosCursosSemestreB[t]: cruces[i.c1][i.c2]> 0) w[i]*cruces[i.c1][i.c2]<= d[t];<br />
    //Nueva: Cuenta Alumnos con pruebas en dias consecutivos
  forall (i in CursosCursos, d in Dias: ord(Dias, d) < 19)<br />        x[i.c1][d] + x[i.c1][next(Dias, d)] + x[i.c2][d]+ x[i.c2][next(Dias, d)]<=1 + l[i];  <br />       
    sum (i in CursosCursos) l[i]*cruces[i.c1][i.c2]<= m;<br />    sum (i in CursosCursos) l[i]*cruces[i.c1][i.c2]*creditos[i.c1][i.c2]<=c;<br />
  //Suma de diff de horas entre pares de pruebas/  forall (i in CursosCursos, d in Dias: ord(Dias, d) < 19 && cruces&#91;i.c1&#93;&#91;i.c2&#93;> 0)
    Ct5: 
    ((x[i.c1][d]/(-horashoras[i.c1][d]+24+horashoras[i.c2][next(Dias,d)])) + x[i.c2][next(Dias,d)]/(-horashoras[i.c1][d]+24+horashoras[i.c2][next(Dias,d)]))  <= g[i] +(1-l[i])*999999; <br />
  sum (i in CursosCursos) 6*(g[i])*cruces[i.c1][i.c2]<= h;<br />
forall (i in CursosCursos, d in Dias: ord(Dias, d) < 19 && cruces&#91;i.c1&#93;&#91;i.c2&#93;> 0)
      (-horashoras[i.c1][d]+24+horashoras[i.c2][next(Dias,d)])  <= horas[i] +(1-l[i])*999999;<br />      sum (i in CursosCursos) horas[i]<=horastotales;<br />
forall(i in CursosCursos)
g[i]>=0;
forall(i in CursosCursos)
horas[i]>=0;
};

I saw somewhere that when namignt eh constraints i would need to name it like CT[x]; and then CT[x].duals or something like that, but i keep getting so many problems.

Please help me out..

Thank you very much
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[alain.chabrier said:]

Hello,

indeed you have lots of integere variables in your model.
e.g.

dvar int d[Semestres]; //Cuenta de Alumnos de un mismo semestre con exámenes en días sucesivos
dvar int m; //Cuenta de Alumnos con exámenes en días sucesivos
dvar int c;
dvar int horas[CursosCursos];
dvar boolean w[CursosCursosSemestre]; //si dos pares de curos de un mismo semestre tiene examen en dias consecutivos

With integer variables, you will get a MIP (Mixed Integer Problem) for which you cannot get any dual variables or reduced costs.
This is not a limitation of the product, this is mathematical programming fact.

Alain
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[cwinter said:]

Would i be able to see the duals if i change the integers to floats?. There is only one variable which i cannot change to float, which is W[CursosCursosSemestres]. I can leave this one out of the model. Would that work ok??
#ConstraintProgramming-General
#DecisionOptimization

Originally posted by: SystemAdmin

[Didier Vidal said:]

If you have an LP model, you can see the reduced costs, in OPL 5.5, by right-clicking on the variable in the problem browser (see screenshot).

However, changing integers to floats might change the meaning of your model.... Especially if you have boolean variables.
#ConstraintProgramming-General
#DecisionOptimization