Machine learning for modelling and identification of educational robotics activities

Educational Robotics (ER) is a powerful tool to help students learn school subjects, robotics, and developing cognitive skills and soft skills. Assessing the learning outcomes of ER activities requires the identification of the model that underly the process. Machine learning can be useful to identify such models and to interpret data. This paper aims to present a system that could help integrating Educational Data Mining and Learning Analytics techniques into the open-ended learning environment that characterizes the constructionist approach of ER. Both supervised and unsupervised learning methods could be applied to extract meaningful information. Students' approaches to learning as well as a prediction of their final performance could inform teachers' decision and facilitate the implementation of effective ER activities in formal and non-formal education. First results show good premises for a future broader implementation, but more research is needed to face all the open issues.