Otimização da função de avaliação em algoritmos genéticos para CBCTT : redução de inversão de prioridades na evolução
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Publikationsdatum: | 2025 |
| Format: | Bachelor thesis |
| Sprache: | por |
| Quelle: | Repositório Institucional da UPF |
| Download full: | https://repositorio.upf.br/handle/123456789/9867 |
Zusammenfassung: | The construction of academic timetables is a complex task involving multiple hard and soft constraints, whose violations directly affect the quality of the generated solutions. The genetic algorithm currently used by the University of Passo Fundo relies on a weighted evaluation function, but its formulation allows priority inversions, where improvements in low-importance constraints compensate inappropriate deteriorations in critical ones. This study introduces a redesigned evaluation function based on an explicit separation between hard and soft constraints, combined with a hierarchical ordering criterion for ranking individuals. Experiments were conducted using two institutional datasets, totaling 192 independent executions. The results demonstrate that the proposed approach completely eliminates priority inversions while preserving performance levels regarding hard constraint violations when compared to the model with weights and limits. Furthermore, the reformulated evaluation function reduces sensitivity to manual adjustments of weights and limits, simplifying maintenance across academic terms. The findings indicate that the proposed model leads to a more coherent and predictable evolutionary process, better aligned with institutional requirements, and contributes to greater stability and reliability in timetable generation. |