Benchmark Dataset for the Multi-Layer Anaesthetist Rostering Problem
Description
This dataset provides a comprehensive benchmark for the Anaesthetist Rostering Problem (ARP), a multi-layer healthcare workforce scheduling problem involving monthly on-call, weekly daytime, and operating theatre room assignments. The dataset includes 32 synthetic instances across four sizes (Small, Medium, Large, XLarge) and four difficulty levels (Easy, Medium, Hard, VeryHard), along with 5 months of anonymised real-world operational data from a Malaysian teaching hospital. The constraint model comprises 8 hard constraints and 21 soft constraints with dual-matrix fairness enforcement, all verified against real-world data with zero hard constraint violations. Baseline results from IBM CPLEX 22.1 and Google OR-Tools CP-SAT 9.12 are provided for all instances, together with an evaluated manual roster for comparison. Source code for data loading, instance generation, solver implementation, and manual roster evaluation is included to support reproducibility.
Files
Steps to reproduce
1. Extract the ZIP file 2. Install IBM CPLEX 22.1 (academic license available at ibm.com/academic) and/or Google OR-Tools 9.12 3. Compile Java source code: javac -cp "lib/" -d build src/main/java/com/arp/**/.java 4. Run CPLEX solver on real-world data: java -cp "build;lib/*" -Djava.library.path="" com.arp.solver.CPLEXRunnerV2 data/real-world/layer1 5. Run OR-Tools solver on synthetic instance: java -cp "build;lib/*" -Djava.library.path=lib com.arp.solver.ORToolsRunnerV2 data/synthetic/S_E_01 6. Evaluate manual roster: java -cp "build;lib/*" com.arp.evaluator.ManualRosterEvaluatorV2 1 5 7. Results will be generated in output/cplex/, output/ortools/, and output/manual/ folders
Institutions
- National University of MalaysiaSelangor, Bangi