Improving the ILS-TQ technique for the high school timetabling problem

Abstract

The High School Timetabling Problem is an NP-Complete problem that consists in allocating subjects, that are taught by teachers and assigned to each class, to periods while satisfying constraints. Throughout the years, metaheuristics haven given better results to real-life instances compared to deterministic methods since the search space for timetabling problems are huge and exploring it completely is impossible. The better the schedules are, the better the students and teachers’ performance, and the costs of generating these schedules are reduced. This proposal consists in modifications done separately to the Iterated Local Search (ILS) with the Torque (TQ) operator for the 34 real-life instances of schools of Brazil. These separate modifications change how a schedule is modified and how it is is accepted. Our Simulated Annealing (SA) cooling scheme implementation, with some parameter tuning, gave better results than our other methods, and more consistent solutions than the original method for some instances. Furthermore, to create other instances more easily, a form was created.