Aspects of computerised timetabling

This research considers the problem of constructing high school timetables using a computer. In the majority of high schools, termly or yearly timetables are still being produced manually. Constructing a timetable is a hard and time consuming task which is carried out repeatedly thus a computer program for assisting with this problem would be of great value. This study is in three parts. First. an overall analysis of the problem is undertaken to provide background knowledge and to identify basic principles in the construction of a school timetable. The characteristics of timetabling problems are identified and the necessary data for the construction of a timetable is identified. The first part ends with the production of a heuristic model for generating an initial solution that satisfies all the hard constraints embodied in the curriculum requirements. The second stage of the research is devoted to designing a heuristic model for solving a timetable problem with hard and medium constraints. These include constraints like the various numbers of common periods, double periods and reducing the repeated allocation of a subject within any day. The approaches taken are based on two recently developed techniques, namely tabu search and simulated annealing. Both of these are used and comparisons of their efficiency are provided. The comparison is based on the percentage fulfilment of the hard and medium requirements. The third part is devoted to one of the most difficult areas in timetable construction, that is the softer requirements which are specific to particular schools and whose satisfaction is not seen as essential. This section describes the development of an expert system based on heuristic production rules to satisfy a range of soft requirements. The soft requirements are studied and recorded as rules and a heuristic solution is produced for each of the general requirements. Different levels of rule are developed, from which the best possible solution to a particular timetable problem is expertly produced. Finally, possible extensions of the proposed method and its application to other types of the timetabling problem are discussed.