The goal of the work is to improve the teaching-learning process through the inclusion of prediction features in a control system proposal namely Reactive Blended Learning. To achieve this goal, a model of the student has been proposed, whose considered outputs are the performance and a participation index that measures the activity level of the student in the class. The controller is based on fuzzy logic and uses the predictions of the model to anticipate the student's state. An important issue that has been taken into account is the limited time to identify the dynamics of the student learning before the course ends. This limitation has been treated through a three-stage process. It is important to remark that this work is not focused on obtaining a complete student model, but on getting useful information for the detection of trends in the teaching-learning process. Preliminary results on a real course are presented to attest the efficiency of the proposed control strategy.