Novel Knowledge Based and Heuristic Approaches to University Timetabling

The research presented in this PhD thesis contributes to a larger project to develop a case-based reasoning (CBR) meta-heuristic selector for examination timetabling problems. Chapter 2 gives a detailed survey of techniques in the literature applied to examination timetabling. The work in this thesis aims to provide a significant contribution to the examination timetabling domain in its own right, as well as contributing to work on the CBR system. The motivation behind a CBR system is that similar problems can be solved equally well by a similar technique. One of the main aspects of such a case based reasoning system are how two problems can be measured as similar in such a way that this assumption holds. Chapters 4, 5 and 6 investigate how a measure of similarity between two examination timetabling problems can be developed. The key features of a number of benchmark problems are examined and the behaviour of the problems when optimised using simulated annealing is also considered. This enables a far greater understanding of the problem structure as well as determining strong features by which to measure similarity. Chapter 7 gives a brief description of the CBR system developed in the wider project. The other key aspect of this thesis is the development of a generalised metaheuristic to competitively solve examination timetabling problems. Chapter 8 introduces a variable neighbourhood search (VNS) technique which proves to be highly successful on a number of benchmark problems as well as providing a high degree of flexibility. This flexibility is further exploited in Chapter 9 where a variant of VNS is proposed which uses a genetic algorithm to evolve the best set of neighbourhoods for a given problem. This technique proves highly competitive across a range of problems by itself, but can also be used to provide the basis for a CBR system, as discussed in Chapters 9 and 10.