<![CDATA[Course scheduling is a complex problem in higher education because it must satisfy multiple constraints involving courses, instructors, rooms, and time slots. This study examines the impact of population size variation on the computational efficiency of a Genetic Algorithm (GA) applied to a medium-scale instance consisting of 35 courses, 15 instructors, 12 rooms, and 20 time slots. Simulations were conducted in MATLAB using population sizes ranging from 20 to 1000, while all other GA parameters were held constant to isolate the effect of population size. Solution quality was evaluated using a conflict-based fitness function, and all configurations yielded valid timetables with zero hard-constraint violations. Experimental results reveal a consistent non-linear relationship between population size and computation time. Statistical findings in Table 1—including mean values, standard deviations, and 95% confidence intervals—show that both very small and very large populations produce higher and more variable execution times. In contrast, population sizes of 300–400 achieve the lowest and most stable computation times, indicated by the smallest mean values and narrow confidence intervals. For the instance and configuration used in this study, this range serves as an effective starting point for population size tuning. Overall, the findings highlight the importance of empirical parameter selection to balance computational efficiency and solution quality in academic timetabling systems.]]>
Copyrights © 2026
