Article ID: | iaor20104805 |
Volume: | 40 |
Issue: | 3 |
Start Page Number: | 196 |
End Page Number: | 207 |
Publication Date: | May 2010 |
Journal: | Interfaces |
Authors: | Miranda Jaime |
Keywords: | timetabling |
Each October, the Executive Education Unit at the Universidad de Chile develops its course schedules for the following year. By 2008, the complexities of increasing enrollments and course offerings had rendered its manual timetabling process unmanageable. Inconvenient and inflexible scheduling decisions were causing discontent among instructors and students, making the need for a more efficient system of assigning classrooms patent. Three characteristics distinguish the unit's situation from the classic university course timetabling problem. First, its courses vary in duration, ranging between 15 and 30 weeks. Second, its course start dates are spread over the academic year. Finally, each course's start date is flexible and must fall within a window defined by the earliest and latest start dates. This paper presents an automated computational system that generates optimal timetables and classroom assignments for all the unit's courses, minimizing both operating costs and schedule conflicts. When we compared the schedules it generated with the unit's manually generated timetables, we found that our system yielded average cost savings of 35 percent; in addition, it reduced execution times (for generating schedules) from two weeks to less than 30 minutes.