Diversity in Genetic Algorithms in the Generation of School Schedules

The Timetabling task is classified as a NP-complete problem; therefore, several constrains must be adjusted to generate valid timetables. In this type of problem, it is well known that the complexity of this task grows exponentially with the increasing number of variables and range of values, making it unfeasible to design schedules manually. Given their effectiveness in tackling large search spaces, Genetic Algorithms (GA) have emerged as a promising tool for addressing the Timetabling problem. Thus, an important aspect of GA is to maintain diversity among individuals during evolution, aiming to converge to an optimal solution efficiently. Therefore, this study focuses on exploring similarity techniques to measure diversity in order to improve GA individuals in timetabling generation. The obtained results demonstrate the population's evaluation performance, indicating higher accuracy with Jaccard similarity and faster evaluation with Hamming distance.