Enabling Mixed Genetic Algorithm for Automatic Group Formation System

Group formation plays a crucial role in designing collaborative learning activities, as the success of the group largely relies on the makeup of its members. While numerous algorithms exist, many group formation systems tend to adopt a single grouping strategy, such as either heterogeneous or homogeneous grouping, limiting their ability to address diverse student characteristics simultaneously. In this paper, we propose an integrated approach utilizing a mixed genetic algorithm within a data-driven learning platform, which considers both homogeneous and heterogeneous characteristics concurrently. Through an exploratory implementation in a university course, we examined the algorithm’s performance using authentic log data under various grouping strategies in classroom settings. We also highlight the potential of interpretability in group formation results, particularly through the composition panel, enabling teachers to make informed interventions and thereby enhancing overall class performance.