Learning analytics to monitor and predict student learning processes in problem solving activities during an online training

Research on Learning Analytics is closely related to the use of a Digital Learning Environment, which can be defined as a learning ecosystem in which to teach, learn and develop skills in the classroom, online or in hybrid mode. Using this, Educational Data, continuously updated and growing, have become Big Data. To make the most of these data, it is useful to use Learning Analytics techniques to analyze and interpret them, and to obtain enough information to make decisions. Predicting student learning outcomes is one of the main topics in learning analytics research. This research work has the main goal of building a model for predicting which score range students will achieve at the end of the online training. In this way, in future editions of the project, by analyzing the situation of students during the online training, we may provide students with personalized feedback to increase their involvement in the training and prevent dropouts. We analyzed three past editions of the Digital Math Training online training and developed two Random Forest models capable of predicting the final score range obtained by the students after the first three and the first six problems. We have also developed two Probabilistic Neural Network models with the same purpose, but with worse results.