Particle swarm optimization for solving thesis defense timetabling problem

Abstract

The thesis defense timetabling problem is a fascinating and original NP-hard optimization problem. The problem involves assigning the participants to defense sessions, composing the relevant committees, satisfying the constraints, and optimizing the objectives. This study defines the problem formulation that applies to Universitas Multimedia Nusantara (UMN) and use the particle swarm optimization (PSO) algorithm to solve it. As a demonstration of concept and viability, the proposed method is implemented in a web-based platform using Python and Flask. The implementation is tested and evaluated using real-world instances. The results show that the fastest timetable generation is 0.18 seconds, and the slowest is 21.88 minutes for 25 students and 18 department members, without any violation of the hard constraints. The overall score of the EUCS evaluation for the application is 4.3 out of 6.

Priority Scheduling Implementation for Exam Schedule

“…Some research is trying to improve this algorithm [10]. The particle swarm optimization (PSO) algorithm can be split into three steps: the first step is to initialize each particle's position and velocity, the next is to update the velocity, and lastly is to update the particle's position [11].…”

A Multi-Objective Model for Thesis Defence Scheduling

“…Conversely, Battistutta et al (2019) is the first to impose an exponential penalty for the number of times a committee member is assigned to a different committee. Finally, Christopher and Wicaksana (2021) includes a maximum quota of assignments for each committee member.…”

“…In terms of the methodologies used to assess committee suitability, there is minimal variance. Each consideration of this point of view aims to optimise the matching between expertise or research areas of the committee members and the subject of the defences they are assigned to (Huynh et al, 2012;Pham et al, 2015;Battistutta et al, 2019;Christopher and Wicaksana, 2021). Nonetheless, other aspects are also regarded in Christopher and Wicaksana (2021), specifically committee member academic level and previous experience moderating defences.…”

“…There are several types of defence scheduling problems, distinguished by the number of defences that a committee is assigned to: (1) Single defence assignment -Each committee is assigned to one defence, which will take place in one slot (day, hour and room). This type of scheduling is addressed in Huynh et al (2012); Pham et al (2015); Limanto et al (2019); Tawakkal and Suyanto (2020); Christopher and Wicaksana (2021) and our work; (2) Session of defences assignment -Each committee is assigned to a group of defences (session), which will take place in one slot (day, period and room). Evidently, the time slot cannot be just a particular hour in such instances.…”

“…Several different solution approaches have been applied to the thesis defence scheduling problem, specifically: Mixed-integer linear programming (Battistutta et al, 2019); Constraint programming (Battistutta et al, 2019); Greedy backtracking hybrid algorithm (Su et al, 2020); Local search (Kochaniková and Rudová, 2013;Pham et al, 2015;Battistutta et al, 2019;Tawakkal and Suyanto, 2020); Genetic algorithm (Huynh et al, 2012;Limanto et al, 2019); and Particle swarm optimisation (Christopher and Wicaksana, 2021).…”