Zheng, ZhilinPinkwart, NielsIgel, ChristophUllrich, CarstenWessner Martin2017-10-052017-10-052017978-3-88579-667-1https://dl.gi.de/handle/20.500.12116/4857Group re-composition has thus far been rarely studied. The recent emergence of large scale online learning contexts (e.g. MOOCs) might bring about an opportunity for its application due to the reported high drop-out rate. In this paper, we propose a novel data-driven approach to address the problem of group re-composition. Through a simulation experiment, we saw its capability in decreasing the drop-out rate in groups and bringing more cohesive groups when compared against a random grouping strategy.enGroup FormationGroup Re-compositionGroup DynamicsMOOCLearning Analytics.1617-5468