Efficient Academic Scheduling at the U.S. Naval Academy

Abstract

Abstract : This research project examined academic scheduling problems at the U.S. Naval Academy. The focus was on devising methods to construct good final exam schedules and improve existing course schedules by facilitation course changes. The final exam scheduling problem is an example of an NP-hard problem. These difficult problems do not admit efficient deterministic solutions. Several heuristic methods to treat these problems were considered. An approach using genetic algorithms showed particular promise. Genetic algorithms involve mating parent schedules to form favorable offspring schedules and then subjecting these new schedules to local mutation. A computer program implementing these ideas was created and tested. Section changes at the Naval Academy had been done on an ad-hoc basis, but this project determined that it could be streamlined and improved by using a centralized barter system. The barter technique accepts input listing desired section changes and identifies multi-student section changes to accommodate their desires. A prototype computer program that uses network flow algorithms to find such section changes was devised. In addition, a method incorporating integer programming techniques was examined and tested.