Exploration-Exploitation Balanced Krill Herd Algorithm for Thesis Examination Timetabling

The thesis examination is one of the many events that matter to student graduation. Making a schedule for the session by the coordinator at a related instance is time-consuming. Besides, the balance between intensiveness and fairness should be one concern so that examiners do not have a tight schedule, but the total duration does not take too long. A Krill Herd (KH) algorithm, one of the nature-inspired algorithms, is examined. The basic KH is used for local search. Both genetic operators and elitism are added to balance its explorative-exploitative movements of krill individuals to find the global solution. The explorative movement randomly searches the global solution in all directions, while the exploitative movement just finds it in a potential area. A dataset collected from the Faculty of Informatics, Telkom University in 2017-2018, which contains 25 students and 110 lecturers, is used to evaluate the model. By carefully tuning two parameters, the KH-based model produces a schedule with a few violated constraints with the effectiveness of 92%, which can be acceptable to be used in the real application. This result shows that continuous-type of solving algorithms can solve the discrete-type problem with some adjustments. In the future, a procedure of premature convergence avoidance can be introduced to improve the model.