A multi-objective evolutionary algorithm for examination timetabling
Journal of Scheduling
A multi-agent system for course timetabling
Intelligent Decision Technologies
Hi-index | 0.00 |
This paper describes an automated curriculum timetabling system based on a stochastic search methodology, namely a co-evolutionary algorithm. The application timetable is taken from the undergraduate courses of the School of Electrical and Electronic Engineering (EEE), Nanyang Technological University (NTU). A co-evolutionary algorithm approach is found to be well suited. Practical courses have duration greater than one hour. A schedule can be generated separately and its population, which consists of a set of practical schedules, is termed as the practical population. Lecture and tutorial schedules can also be generated separately. These are of one-hour duration and they are termed collectively as lecture/tutorial schedule. A set of lecture/tutorial schedules could be generated to form the lecture/tutorial population. These two populations use the same set of resources and have constraining effects upon one another. Since the placement of practical courses have a more constraining effect, the schedules in the practical population are first generated and are then used to guide the generation of the set of lecture/tutorial schedules. For every lecture/tutorial schedule generated, it is combined with its corresponding practical schedule to form a combined schedule. The average fitness of all the combined schedules is then computed and used as a measure of the fitness of the practical schedule that drives them. The practical population is then evolved progressively to obtain the best practical schedule. It is then used as a base configuration for the rest of the courses to populate and evolve. The resultant system compares favorably to the current manual system.