Some Experiences with Hybrid Genetic Algorithms in Solving the Uncapacitated Examination Timetabling Problem

TL;DR

This paper explores hybrid genetic algorithms with local search components for solving the uncapacitated examination timetabling problem, demonstrating their effectiveness and similar performance to existing methods on benchmark instances.

Contribution

Introduces two hybrid genetic algorithms with novel solution representations and crossover schemes, highlighting the importance of local search in timetabling solutions.

Findings

Hybrid algorithms outperform pure genetic algorithms.

Local search significantly improves solution quality.

Algorithms perform comparably to state-of-the-art methods.

Abstract

This paper provides experimental experiences on two local search hybridized genetic algorithms in solving the uncapacitated examination timetabling problem. The proposed two hybrid algorithms use partition and priority based solution representations which are inspired from successful genetic algorithms proposed for graph coloring and project scheduling problems, respectively. The algorithms use a parametrized saturation degree heuristic hybridized crossover scheme. In the experiments, the algorithms firstly are calibrated with a Design of Experiments approach and then tested on the well-known Toronto benchmark instances. The calibration shows that the hybridization prefers an intensive local search method. The experiments indicate the vitality of local search in the proposed genetic algorithms, however, experiments also show that the hybridization benefits local search as well. Interestingly, although the structures of the two algorithms are not alike, their performances are quite similar to each other and also to other state-of-the-art genetic-type algorithms proposed in the literature.