topic Re: Optmilp: read composed index with missing values in Mathematical Optimization, Discrete-Event Simulation, and OR

Optmilp: read composed index with missing values

asasgdsfa — Tue, 07 Nov 2023 08:23:50 GMT

Hello

May I ask for an advise? I'm trying to read 2D data to proc optmilp and I'm getting error:
"ERROR: The symbol 'student' has no value at line XXX column YYY"

I think, it is because I'm trying to load data where not all combinations of indices are present.

student	class	payment
1	A	1
2	A	2
2	B	3
3	A	4
3	B	5

So here, you can see that there is not combination of studen 1 in class B. Could this be a reason for the error?

Code:

proc optmilp;
	set student;
	set <str> class;
	
	num income {student, class};
	read data work.input into [student class] payment = income;
	
	...
quit;

I need to have data in the single table and filling missing records (1-B) would probably lead to numerically more difficult task.

I tried to use only a single index (composed of student-class) and it worked, but for the problem, I need to access these two indexes independently.

Re: Optmilp: read composed index with missing values

ChanceTGardener — Tue, 07 Nov 2023 14:01:19 GMT

A couple of suggestions:

1. You're calling the OPTMILP procedure, but the syntax more closely resembles the OPTMODEL procedure. Change PROC OPTMILP to PROC OPTMODEL

*proc optmilp;

proc optmodel;

2. Declare a set of ARCS indexed by both students and classes. Declare a payment parameter indexed by ARCS, and read the table into the OPTMODEL procedure.

set <num,str> ARCS;

num payment{ARCS};

read data work.input into ARCS = [student class] payment;

3. To create a set of STUDENTS and a set of CLASSES from the ARCS set, use the setof{ } operator.

set STUDENTS = setof{<s,c> in ARCS} s;

set CLASSES = setof{<s,c> in ARCS} c;

Full code below:

data work.input;
 input student class $ payment;
datalines;
1 A 1
2 A 2
2 B 3
3 A 4
3 B 5
;

*proc optmilp;
proc optmodel;

set <num,str> ARCS;

num payment{ARCS};

read data work.input into ARCS = [student class] payment;

set STUDENTS = setof{<s,c> in ARCS} s;
set CLASSES = setof{<s,c> in ARCS} c;

/* write the elements of each set to the log */
put ARCS=;
put STUDENTS=;
put CLASSES=;

/* print payment to the results viewer */
print payment;

Structuring it this way allows you to sum across all students, or sum across all classes later in the program without the missing element <1,B> causing an issue:

num total_by_class{c in CLASSES} = sum{<s,(c)> in ARCS} payment[s,c];

num total_by_student{s in STUDENTS} = sum{<(s),c> in ARCS} payment[s,c];

print total_by_class;
print total_by_student;

The MISSINGS set below captures all of the student-class combinations not present in your data, and as a result, not part of the new calculated parameters above.

set MISSINGS = (STUDENTS cross CLASSES) diff ARCS; 
put MISSINGS=;

Re: Optmilp: read composed index with missing values

asasgdsfa — Thu, 09 Nov 2023 07:48:58 GMT

Thank you, this is exactly what I needed!

(Your are correct, it's optmodel, but I was thinking about something else, sorry.)