A genetic algorithm application for university exam timetabling: A real-world study

OpenAlex İngilizce

This study presents the Key Academic Scheduling Program (KASP), a genetic algorithm based application for university exam timetabling across multiple departments with shared rooms and proctoring resources. KASP combines permutation encoding, a constructive decoder with bounded repair, and a configurable penalty framework that captures student and staff preferences, including preventing overlaps, limiting same-day concentration and back-to-back exams, discouraging temporal clustering of an instructor’s exams, and balancing proctoring duties. KASP is evaluated on fifteen real institutional instances drawn from three exam periods (Fall midterms 2024, Fall midterms 2025, and Spring make-up 2025) at the same faculty across four departments, comprising both per-department and joint-faculty optimization scenarios with forced assignments and shared resources. Across twelve per-department instances and three independent random seeds per instance, KASP achieves a 22.7% mean reduction in normalized penalty score relative to the manual baseline (Wilcoxon signed-rank p = 0.001, Cliff’s δ = 0.76, winning 11/12 comparisons). On three joint-faculty instances, KASP achieves a 14.7% mean improvement. Under a matched wall-clock budget based on the GA runtime, KASP also outperforms a Late Acceptance Hill Climbing baseline on every comparison (p < 0.001, Cliff’s δ = 0.93, mean improvement 14.8%). Relative to the manual timetable on the principal instance, KASP removes all cross-class overlaps, reduces the maximum same-day student load to 240 minutes, and yields more even proctoring patterns while maintaining classroom sufficiency. KASP’s Excel-based data exchange and reproducible exports support deployment and auditing.