Suppose I have three employees and I have to assign three employees based on their ranks. If an employee has rank 1 that means he is best. Say, I have the following table

Name      Rank
Joe        3
Allen      1
Bob        2

I have make one parameter $p_{e, rank}.$ If I can not assign it based on rank then I have to give some penalty. This is a part of an employee allocation problem.

What I did, I make a binary variable $x_{e, r, rank}$ for Assignment of employees to restaurant $r.$ Then I create another variable $\sum(x_{e, r, rank} \cdot p_{e, rank}).$ I am minimizing this sum and this is my objective function.

My query is, is it correct? Or if there is any alternative way to write it in a mathematical program let me know.

  • $\begingroup$ Minimizing the sum is a good starting point. One disadvantage is that it may penalize a single person or restaurant quite heavily. It is an interesting exercise to maintain some equality. $\endgroup$ Jun 16 '21 at 1:09

A) I think you might want to use a multi-level score (something like our BendableScore):

  • SOFT: maximize rank 1 assignments
  • SOFTER: maximize rank 2 assignments
  • SOFTEST: maximize rank 3 assignments

Otherwise, three rank 2 assignments can be considered better than one rank 1 assignments (which I presume is not the requirement).

Where it gets more interesting is if your number or ranks is dynamic: One dataset can have 3 ranks, while another could have 3000 ranks. In that case, a fairness like squaring technique probably works too, to collapse it into a single score level.

B) If, on the other hand, outweighing is the requirement, just use a single score level (SOFT) with weighting (one weight between every two sequential ranks). In your current implementation: rank 1 is twice rank 2. Rank 2 is 1.5 times rank 3. That's weird. It would be more consistent weighting wise, to use rank 1 (penalty 1), rank 2 (penalty 2), rank 3 (penalty 4), rank 4 (penalty 8) and so forth.

  • $\begingroup$ Can you suggest a formulation if we use consistent weighting, to use rank 1 (penalty 1), rank 2 (penalty 2), rank 3 (penalty 4), rank 4 (penalty 8) and so forth. @Geoffrey $\endgroup$
    – Manglu
    Jun 16 '21 at 3:49
  • $\begingroup$ It depends on the solver of course, but something like 2^(rank - 1) so penalize("ranking constraint", SOFT, assignment -> 2^(assignment.getRank() - 1)) should get you those weights. Basically 2^0 is 1, 2^1 is 2, 2^2 is 4, etc. $\endgroup$ Jun 16 '21 at 15:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.