Network-Aware Post-Timetabling Examination Room Allocation with Screened-Pool Repair

Abstract

Universities often fix the examination timetable before finalizing room allocation. We start from this fixed timetable and an incumbent room allocation; examination sessions are not rescheduled. We study the downstream post-timetabling examination-room repair problem on real anonymized data from a large public university, covering 81 examination sessions over 19 days, 115,641 candidates, and 3,392 incumbent room assignments. The incumbent plan uses 38.1% of active capacity and has 39.4% mean-room physical utilization. We formulate the task as bipartite network repair with two operational regimes: Legacy-room repair , which reassigns packages among incumbent active rooms, and Screened-pool repair , which expands the candidate-room set with a small controlled pool of unused rooms. We compare a weighted ILP-based policy with transparent repair heuristics. The ILP-family policy reaches 57.6% room reduction under Legacy-room repair and 62.1% under Screened-pool repair, with solver status reported explicitly. Screened greedy reaches 45.3%, robust repair trades savings for higher stress-test feasibility, and adaptive repair recovers 41.8% reduction with millisecond-scale median runtime. We complement the deterministic institutional evaluation with controlled attendance-buffer and room-loss stress tests, which assess feasibility under perturbed assumptions rather than probabilistic attendance forecasts, and with a synthetic-room public-benchmark cross-check. The contribution is an applied, operational repair framework and artifact-driven evaluation for fixed-timetable examination room allocation, rather than a new general-purpose optimization algorithm. Keywords: Examination room allocation, Educational timetabling, Integer linear programming, Greedy matheuristics, Computational networks

Cite this paper

@inproceedings{le2026screenedpool,
  author    = {Le, Minh Hieu and Nguyen, Phung Trung Nguyen and Mai, Duc Thuan and Nguyen, Ngoc Khanh Tran and Le, Xuan-Bach},
  title     = {Network-Aware Post-Timetabling Examination Room Allocation with Screened-Pool Repair},
  booktitle = {Proceedings of the 15th International Conference on Computational Science and Network Intelligence (CSoNet 2026)},
  year      = {2026},
  series    = {Lecture Notes in Computer Science},
  publisher = {Springer},
  address   = {Ho Chi Minh City, Vietnam},
  note      = {Artifact: \url{https://github.com/hieukendu/CSonet}}
}