An application of genetic algorithms to the school timetabling problem

Abstract

There has been a large amount of research into the automatic generation of school timetables. Methodologies such as constraint programming, simulated annealing, Tabu search and genetic algorithms have been applied to the school timetabling problem. However, a majority of these studies focus on solving the problem for a particular school and there is very little research into the comparison of the performance of different techniques in solving the school timetabling problem.The study presented in this paper evaluates genetic algorithms (GAs) for the purpose of inducing school timetables. For each problem, the GA implemented iteratively refines an initial population of school timetables using mutation to find a good quality feasible timetable. The performance of the GA on a set of five benchmark problems has been compared to the performance of neural networks, simulated annealing, Tabu search, and greedy search on the same set of problems. The results obtained by the GA were found to be comparable to and an improvement on those produced by the other methods.