Publication Details

Garshasbi, S., Mohammadi, Y., Graf, S., Garshasbi, S. & Shen, J. (2019). Optimal Learning Group Formation: A Multi-objective Heuristic Search Strategy for Enhancing Inter-group Homogeneity and Intra-group Heterogeneity. Expert Systems with Applications, 118 506-521.


In modern education systems, plenty of research suggests that clustering the learners into optimal learning groups based on their multiple characteristics is a determining effort in enhancing the effectiveness of collaborative learning. Although there have been several evidences on developing and implementing appropriate computational tools to handle classification processes in expert and intelligent systems, the effectiveness and accuracy of optimal grouping algorithms are still worth improving. For instance, the majority of grouping processes in collaborative learning environments is orchestrated through single-objective optimization algorithms, which need to be revisited due to some intrinsic limitations. In this paper, we propose a novel algorithm capable of properly addressing a variety of optimization problems in optimal learning group formation processes. To this end, a multi-objective version of Genetic Algorithms, i.e. Non-dominated Sorting Genetic Algorithm, NSGA-II, was successfully implemented and applied to improve the performance and accuracy of optimally formed learning groups. In contrast to the previous related works applying single-objective algorithms, the main advantage of our work is simultaneous satisfaction of multiple targets predefined for the formation of optimal learning groups, especially the inter-homogeneity and intra-heterogeneity of each learning group, which significantly enhance both effectiveness and accuracy of optimal grouping processes in the underlying intelligent systems. Challenging the proposed optimization algorithms, both single- and multi-objective optimizers, with a similar grouping problem, clearly proved that the single-objective optimization technique has limited control and sensitivity to the quality of individual groups. Contrary to single-objective optimization techniques, which are mainly governed by adjusting the quality of the groups altogether in average, the proposed multi-objective algorithm not only takes the average desirability of all formed groups into account but also precisely monitors the fitness of each group in a potential solution distinctively. The generality of the proposed algorithm makes it a suitable candidate not only to handle optimal grouping in learning environments but also to be competent enough for grouping problems in other domains as well.

Available for download on Saturday, October 17, 2020



Link to publisher version (DOI)