Optimization of Exam Timetabling Using a Genetic Algorithm

Abstract

This paper presents a hybrid method for generating university examination timetables by combining a population-based genetic algorithm with a local improvement stage based on simulated annealing. The hybrid approach balances global exploration and local intensification: a carefully designed encoding and initialization reduces early-stage conflicts, several crossover and mutation operators maintain diversity, and a targeted annealing phase removes residual overlaps and improves soft-constraint satisfaction. The paper presents the problem formulation, the fitness and penalty model that captures realistic institutional constraints, implementation details, and an experimental evaluation on datasets modeled after a large university exam session. Results demonstrate consistent conflict reduction, improved day-to-day balance for student workloads, and robust convergence compared to a baseline genetic algorithm. The method can be integrated with campus scheduling systems and adapted to diverse institutional rules.