Using multimodal data to find patterns in student presentations

Multimodal Learning Analytics is a subfield of Learning Analytics that uses data coming from complex learning environments and collected through alternative devices that are different from those normally observed in the Learning Analytics literature. The present work uses data captured by Microsoft Kinect and organized with Lelikëlen system to find patterns in students oral presentations during a given discipline. For that, a total of 16 different features related to the records of 43 students presentations (85 observations) were used to generate clusters of students with similar behavior. Initial results indicate three main different profiles of students according to their patterns in oral presentations: active, passive, and semi-active. Such findings can be further implemented in Lelikëlen system in order to allow instant feedback to students. Future work will also evaluate how students oral presentations patterns evolve during the semester, and compare patterns of students presentations across areas to evaluate whether there are similarities or not.