Variants and parameters investigations of particle swarm optimisation for solving course timetabling problems

Abstract

University course timetabling problem (UCTP) is well known to be Non-deterministic Polynomial (NP)-hard problem, in which the amount of computational time required to find the optimal solutions increases exponentially with problem size. Solving the UCTP manually with/without course timetabling tool is extremely difficult and time consuming. A particle swarm optimisation based timetabling (PSOT) tool has been developed in order to solve the real-world datasets of the UCTP. The conventional particle swarm optimisation (PSO), the standard particle swarm optimisation (SPSO), and the Maurice Clerc particle swarm optimisation (MCPSO) were embedded in the PSOT program for optimising the desirable objective function. The analysis of variance on the computational results indicated that both main effect and interactions were statistically significant with a 95% confidence interval. The MCPSO outperformed the other variants of PSO for most datasets whilst the computational times required by all variants were moderately difference.