<![CDATA[Course scheduling in vocational high schools (SMK) constitutes a complex combinatorial optimization problem involving multiple hard and soft constraints related to teacher availability, class allocation, and time-slot distribution. Although Genetic Algorithms (GA) have been extensively applied in educational timetabling, existing studies largely emphasize standalone optimization or desktop-based solutions, with limited analytical evaluation of refinement strategies and system-level applicability. This study addresses this gap by empirically evaluating a hybrid GA–Local Search (LS) approach embedded within a web-based scheduling framework. GA is utilized as a global search mechanism to generate feasible schedules that satisfy all hard constraints, while LS is applied as a post-optimization phase to improve solution quality by reducing soft constraint violations. Experiments were conducted using real scheduling data from SMK Yadika 13 Bekasi, involving 3 subjects, 3 teachers, 4 classes, and 12 time slots within a single-day scenario. Although limited in scale, this configuration was deliberately selected to enable transparent analysis of the optimization dynamics and refinement impact of the proposed hybrid approach. The results show that the pure GA produces five soft constraint violations, mainly due to suboptimal placement of cognitively demanding subjects and uneven subject distribution. After applying LS, violations were reduced to two cases, with the fitness value improving from 0.873 to 0.946 and only a marginal increase in computation time (5–7 seconds). These findings demonstrate that local refinement significantly enhances schedule quality beyond conflict-free feasibility. This study contributes scientifically by providing an empirical assessment of GA–LS hybridization for soft-constraint optimization and by establishing a scalable web-based framework that supports future extensions to full-week scheduling and adaptive academic systems]]>
Copyrights © 2025
