Rational and personalized team formation can greatly enhance the efficiency of online collaborative learning. However, in the process of forming learning teams and completing learning tasks, online synchronous communication among team members is often overlooked, and the abilities of team members may change during task execution, which may have an effect on the completion of team tasks. To address these problems, this paper comprehensively considers the impact of learners’ online time and skill levels on team formation. We quantify the potential development space of learners’ skills and design optimization objectives accordingly based on the Zone of Proximal Development theory. Additionally, we propose a Near-Informer deep learning model to predict learners’ future online states, aiming to measure communication efficiency. Then, we propose a multi-task oriented multi-objective optimization team formation problem, and transform it into a single-objective one by using weighted aggregation of all objectives. After that, we present a Three-dimensional Monte Carlo Tree Search algorithm (TDMCTS). The experiments are conducted based on two real datasets. Compared with the baselines, our Near-informer model achieved the highest number of optimal predictions, with 22 out of 32. Besides, TDMCTS algorithm achieved more comprehensive team formation results, with average skill satisfaction degrees of 99.8% and 98.9% on two task sets, respectively, while satisfying the task requirements.
Read full abstract