Originally posted by: pdst12005

Hello everyone.

I have a huge matrix of decision variables (binary). To each variable there is a date associated thus I would like to run the optimization for each day (it is 6 years so I would need it automated).

I though I could maybe use a for all cicle, e.g:

Xij (1 if record i connects with record j, 0 if not)

Di (date of records i)

Dj (date of records j)

For all Di=Dj

    Max Sum Xij*

    constrains:

    (...)

end for all.

But I am not sure if this is possible, or if there is a simpler way. For instances, programing several optimizations, one for each day.

Thanks in advance

Originally posted by: davidoff

Look at the mulprod sample in the opl examples that illustrate how you change the model and relaunch successive optimization. Basically, there are two ways to do this in OPL,

1) change the data, regenerate a new model from the same definition and an updated data elements.

2) change the coefficients or bounds directly in the extracted model and re run

Hope this helps

David

Originally posted by: pdst12005

Thanks. I though of this but the problem is that I have 365*6 days to individualy optimize thus I wondered if there was a way to automaticaly do this from the main data.

Originally posted by: fbahr

Basically... what @davidoff already said -- just adding an OPL script example.

Assuming you had your model in "model.mod", general data definitions in "data.dat", and day-specific inputs in files "day1.dat" .. "day7.dat", you would create a new .mod file, say main.mod, with:

main
{
    var cplex       = new IloCplex();
    var data_common = new IloOplDataSource("data.dat");
    var model_def   = new IloOplModelDefinition(new IloOplModelSource("model.mod"));
    var oplmodel;
    for (var run = 1; run <= 7; run++)
    {
        cplex.clearModel();
        oplmodel = new IloOplModel(model_def,cplex);
        oplmodel.addDataSource(data_common);
        oplmodel.addDataSource(new IloOplDataSource("day" + run + ".dat"));
        oplmodel.generate();
        if (cplex.solve())
            writeln(cplex.getBestObjValue());
        oplmodel.end();
    }
    // data_common.end();
    // ...
    cplex.end();
}

--fbahr

Originally posted by: pdst12005

And is there a scrip to generate the individual data files?

What I have is a excel sheet with records in each lines and columns would be date and other parameters. So I would need to add a scrip for cplex doing this.

(I could do it using other program but my point is if there is a way to do it with cplex, and reproduce in case I would get a diferent dataset).

Still I am already very thankful for yours and @davidoff help.

additional information:

What I need to do is not so hard. Simply a max function with decision variable being if record i match record j. But I have like 50k reocrds in each i and j sets, thus this would be huge computing amount. What I try to do is to use a blocking variable to reduce the number "comparisons" that is the date (day). This means if day i is different from day j there will be no match for sure. But adding it as constrain would it be enough to reduce the computing?

Either way I will try making the full model and give it some tries and consider your solution by using just a simple exemple with few days.

Originally posted by: fbahr

Sub SplitData()
    Application.ScreenUpdating = False

    Dim Days As Collection
    Set Days = New Collection
    Dim WSID As Integer
    WSID = 1

    Dim SourceSheet, TargetSheet As Worksheet

    Set SourceSheet = Worksheets( ... ) '< worksheet name
    Dim SourceRow, TargetRow As Long
    Const Col As String = "A"
    Const FirstRow As Integer = 1 '< offset
    Dim LastRow As Long

    LastRow = SourceSheet.Cells(SourceSheet.Rows.Count, Col).End(xlUp).Row
    Dim DayStr As String

    For SourceRow = FirstRow To LastRow
        Set TargetSheet = Nothing

        DayStr = SourceSheet.Cells(SourceRow, Col).Value

        On Error Resume Next
        Set TargetSheet = Worksheets("Day_" & Days.Item(DayStr))
        On Error GoTo 0

        If TargetSheet Is Nothing Then
            Days.Add CStr(WSID), DayStr

            Set TargetSheet = Worksheets.Add(After:=Worksheets(Worksheets.Count))
            TargetSheet.Name = "Day_" & CStr(WSID)
            TargetSheet.Range("A1").Value = 0
            WSID = WSID + 1

        End If

        TargetRow = TargetSheet.Cells(TargetSheet.Rows.Count, Col).End(xlUp).Row + 1
        SourceSheet.Rows(SourceRow).Copy Destination:=TargetSheet.Rows(TargetRow)
        TargetSheet.Range("A1").Value = TargetSheet.Range("A1").Value + 1

    Next SourceRow

    Application.ScreenUpdating = True
End Sub

main
{
    var cplex       = new IloCplex();
    var model_def   = new IloOplModelDefinition(new IloOplModelSource("model.mod"));
    var model_dat   = new IloOplDataSource("data.dat");   
    var basic_model = new IloOplModel(model_def,cplex);
    basic_model.addDataSource(model_dat);
    basic_model.generate();
    for (var run = 1; run <= 2190; run++)
    {
        cplex.clearModel();
        var custom_def = basic_model.modelDefinition;
        var custom_dat = basic_model.dataElements;
        var custom_mod = new IloOplModel(custom_def,cplex);
        custom_dat.wsid = "Day_" + run;
        custom_mod.addDataSource(custom_dat);
        custom_mod.generate();
        if (cplex.solve())
            writeln(cplex.getBestObjValue(<wbr>));
        custom_mod.end();
    }
    basic_model.end();
    cplex.end();
}

with [model.mod]

string wsid = "Day_1";
string rows = ...;
...

and [data.dat]

SheetConnection sheet(" ... ");
rows from SheetRead(sheet,wsid + '!A1");
tuples from SheetRead(sheet,"wsid + '!A2:Z" + rows);
...

--fbahr

Originally posted by: pdst12005

Thank you very much for your important help.

Best regards,

Marco Amorim