Update 28-01-2019: We have updated the real life case mix files and corresponding instances. Note that if you have downloaded the benchmark set before 28-01-2019, these files have changed. We will add the updated first solutions and benchmark subset soon.
Here you can find the surgery scheduling benchmark set proposed in Leeftink and Hans (2018) based on a deterministic proximity evaluation. The paper is currently under review. For more information on the paper, please contact Erwin Hans (email@example.com).
The benchmark set consists of diverse instances, both based on theoretical and real-life data. Furthermore, proximity statistics are included for all instances of each parameter combination. In order for researchers to generate instances according to their needs, an instance generator can be found here.
For each instance in the benchmark set, 200 samples are available with the authors. For more information, please contact Erwin Hans (firstname.lastname@example.org). An instance sampler is provided here, in order to generate your own realizations/samples from a specific instance.
All programs were developed in the Embarcadero Delphi programming language, and compiled to MS Windows executables.
We consider two variants of the surgery scheduling problem.
In variant A, planning surgeries in overtime, also known as overbooking, is not allowed. Therefore, an important performance measure of algorithms is the number of cancelled patients.
In variant B, all surgeries in an instance should be scheduled, if necessary in overtime. The planned overtime is an important performance measure.
In both variants, the following constraints apply:
- There can only be one surgery scheduled in one operating room per time unit.
- A surgery cannot be interrupted, and should be finished in the room it was scheduled.
- Surgeries cannot start before the opening time of the operating room.
The performance indicators that are reported upon in the results section are:
The total available capacity minus the sum of total time of scheduled surgeries in regular hours, divided by the number of operating rooms.
# Cancellations (in variant A)
The total number of surgeries not scheduled
Cancellation time (in variant A)
The total time of cancelled surgeries (minutes)
Overtime (in variant B)
Sum of total time of scheduled surgeries in overtime (minutes), divided by the number of operating rooms.
When submitting solutions to the surgery scheduling instances, please use the following format:
- The file should be formatted as a .txt file
- Every row (new sentence) corresponds with the solution of one instance. In this row, the following items are given, tab separated.
- First, the name of the instance is given,
- Then the surgery numbers (space separated) of the first OR are given,
- Followed by the surgery numbers (space separated) of the second OR,
- Followed by the surgery numbers (space separated) of the last OR.
- An example of an input row of the solution file is:
2885_ordays_5_load_0,8_8 1 6 7 2 8 13 10 4 12 20 5 15
- If an OR is empty, please note that sufficient tabs are needed to guarantee your solution to be recognized as a valid solution.
- In one solution files, solutions to multiple instances can be gathered. Preferably, create one solution file per case mix type. This way, for all instances, 29 solution files are to be created.
An example of a solution file for all instances of theoretical case mix 4 can be found here. Before submitting your solutions, please validate your solution with our solution validator, which can be found here.
The best known solutions can be downloaded here. In this solutions folder, one can find separate results for variant A and B. Both the solution files, and the solution output are provided.
- Leeftink A.G. and Hans E.W. (2018). Development of a benchmark set and instance classification system for surgery scheduling. Journal of Scheduling 21(1):17-33.