1939-1382 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TLT.2019.2927914, IEEE Transactions on Learning Technologies TLT-2019-02-0081.R2 1 Abstract— Social Networking-based Learning (SN-Learning) is one of the most promising innovations to promote learning via a social network, and thus, providing a more interactive, student- centered, cooperative and on demand environment. In such an environment, group formation plays an important role to the effectiveness of learning process. Adequate groups foster student interactions and increase learning outcomes. However, group formation is a complex task and requires automatic approaches to produce the optimal results in short time. To this direction, this paper presents a novel genetic algorithm for student grouping in a SN-Learning system. Its innovations pertain to the attributes used for the composition of groups and genetic operators applied. In particular, student attributes refer to the three main dimensions of learning in a SN-Learning environment: academic, cognitive, and social. Regarding genetic operators, the algorithm performs two crossover operators: a modification of 2-point crossover and a new approach, called 1-point per group crossover. Evaluating the proposed algorithm performance, the results show that it is more efficient than simple genetic algorithm approach, and considers a larger number of parameters than usual. Moreover, from pedagogical perspective, a positive students’ attitude and high acceptance towards our group formation method is indicated. Index Terms—Collaborative Learning, Genetic Algorithm, Group Formation, Multi-Criteria Grouping, Social Networks. I. INTRODUCTION HE proliferation of Internet and the emergence of social media technologies have led to the evolution of digital learning, ranging from web-based individual learning environments where students learn individually at their own pace, to Social Networking-based Learning (SN-Learning) ones where students connect to each other via a social network and their interaction constitutes the core of learning process [1]. Social networks (SNs) enable users to create connections with other peers, developing interpersonal relationships, and interact with them without time and place constrictions, through communication and sharing information. Therefore, the exploitation of social networks in education promotes the collaborative learning, providing a more interactive, student- Manuscript received February 27, 2019; revised May 22, 2019. centered, cooperative and on demand environment [2]. The rising SN-Learning environments bring a new dimension to student collaboration. The stimuli of collaboration and student engagement in this process are differentiated using such technology from traditional Computer-Supported Collaborative Learning (CSCL) [3] – [6]. Students can make their friends’ network, communicate with other peers instantly, generate their content and share it, create posts and comment others’ posts, and express their feelings through reaction buttons. Hence, each student can be characterized by a social profile, emerged from this interaction. Student’s social profile plays an important role in collaborative learning and should be considered in group formation process. Group formation is a crucial task in collaborative learning, since the way groups generated and their students’ composition affect the learning outcomes, and accordingly, the efficiency and effectiveness of CSCL environments [7]. The most studies focus on grouping based on students’ profile information and their performance in course [8], [9]. However, the adoption of student social skills as an additional characteristic for group formation can lead to adequate groups in which member interaction is improved, student engagement is fostered, and group potential is increased [10]. In SN-Learning, this characteristic can be easily evaluated by the student interactions in the network. There are four different types of groups in the group formation problem: homogeneous; members with common characteristics, heterogeneous; members with differential characteristics values, mixed; members with some common characteristics and others different, and balanced; members with distributed characteristics i.e. strong students are grouped with weak and average students, and hence, there are no 'strong' or 'weak' groups [8]. The choice of the proper group type depends on learning context and instructional goals [11]. However, the group type is another factor to be considered as it plays an important role in the method used for composing groups. There are different methods to form groups: a. random- selection; groups are generated randomly, b. self-selection; A. Krouska and M. Virvou are with the Department of Informatics, University of Piraeus, 80, M. Karaoli & A. Dimitriou St., 18534 Piraeus, Greece (e-mail: akrouska@unipi.gr; mvirvou@unipi.gr). An Enhanced Genetic Algorithm for Heterogeneous Group Formation based on Multi-Characteristics in Social Networking- based Learning Akrivi Krouska, and Maria Virvou T