Multi-neighborhood local search with room split balancerfor exam timetabling : A case study

This study explicitly addresses the examination timetabling problem (ETP) at University Malaysia Sarawak (UNIMAS), which encompasses both online and physical exams treated within a unified framework of uncapacitated and capacitated formulations. Currently, faculty exam timetabling managed by proprietary systems meets basic constraints but needs to incorporate faculty and stakeholder preferences into a mathematical formulation, making solution quality difficult to assess. To address this issue, we propose a mathematical model that includes university-wide constraints and considers extended soft constraints that accommodate faculty and stakeholder preferences for room sharing and achieving balanced exam splits for shared and non-shared exam scenarios. We introduce a two-stage multi-neighborhood local search method with a balancer to produce high-quality solutions that meet these constraints. Our approach outperforms existing proprietary systems by meeting all standard constraints and achieving extended soft constraints, improving scheduling efficiency and stakeholder satisfaction, and offering a more optimal solution for real-world exam timetabling.