Automated timetabling improves course scheduling at UCLA

Every school must generate a timetable for its courses. If the number of classes is large – say 100 or more – and constraints and preferences are involved, this problem becomes too time-consuming to do manually. I implemented a course timetabling system for the Anderson School of Management at UCLA. I divided the overall problem into two subproblems: first to schedule the core courses using an integer-programming algorithm, and second to generate a timetable by using a heuristic algorithm to solve a variant of a quadratic assignment problem. The administrators have been using this decision support system over the past three years to schedule the courses every quarter and state that it generates timetables as good as the manual ones but a lot faster.