<![CDATA[Examination seating assignment is a combinatorial optimization problem with direct implications for exam assessment’s academic integrity. Existing approaches commonly model constraints in terms of course enrollment adjacency or room capacity, but rarely incorporate student behavioral attributes that proxy for social familiarity and collaboration risk. This study proposes a risk-aware seating formulation in which three correlated student attributes, academic major similarity, enrollment cohort similarity, and registration timestamp proximity, are encoded as weighted pairwise penalty components within a unified fitness function. Three metaheuristic algorithms are implemented and compared: Genetic Algorithm (GA), Simulated Annealing (SA), and the Flood Algorithm (FA). Each algorithm was executed across 30 independent runs on a controlled synthetic dataset of 80 students distributed across 4 examination rooms. Performance was evaluated using descriptive statistics and the Mann-Whitney U test. FA achieved the best mean penalty (103.10) with the lowest standard deviation (1.04), followed by SA (106.03) and GA (110.73). All pairwise differences were statistically significant at α = 0.05. An ablation study further revealed that enrollment cohort similarity is the most impactful constraint parameter, with its inclusion alone sufficient to produce statistically significant algorithmic differentiation. These results demonstrate that FA is the most effective and stable algorithm for this problem formulation, and that registration timestamp proximity constitutes a novel and informative behavioral risk proxy for exam seating optimization.]]>
Copyrights © 2026
