Evaluate the performance of university course timetabling problem with different combinations of genetic algorithm

Foo, Yao Heng (2025) Evaluate the performance of university course timetabling problem with different combinations of genetic algorithm. Final Year Project, UTAR.

PDF Download (3186Kb)

Abstract

University course timetabling problem (UCTP) is a scheduling problem that requires courses to be assigned to the limited time slots, classrooms, and lecturers, while adhering to a set of predefined constraints. Due to the effectiveness of genetic algorithm (GA) in optimisation problems, it has been widely discussed in numerous research to address UCTP. Nonetheless, the performance of GA in terms of operation techniques has not been studied enough, as the researchers have often focused on using a single GA combination or hybrid approaches to solve UCTP case studies. Therefore, this project aims to analyse the performance of different combinations of GA operation techniques and identify the best GA model. A flexible GA framework is developed, which allows alternative techniques to be integrated and executed easily. 64 combinations, involving 4 selection, 4 crossover, 1 mutation, and 4 replacement techniques, are evaluated on a partial mock dataset. In addition, this project proposes a new soft constraint, which requires consecutive classes for a student to be held in the same building. This constraint targets to reduce students’ travel distance, thus producing a more student friendly timetable. Experimental results shows that GA44 model which comprises of binary tournament selection, uniform crossover, swap mutation, and weak chromosome replacement is the best GA combination. In conclusion, the proposed constraint demonstrates clear benefits to student experience on campus and offers a fresh idea for future research with alternative approaches.

Actions (login required)