Teaching Institutions could benefit from Early Warning Systems to identify at-risk students before learning difficulties affect the quality of their acquired knowledge. An Early Warning System can help preemptively identify learners at risk of dropping out by monitoring them and analyzing their traces to promptly react to them so they can continue their learning in the best conditions. This paper presents a novel method for predicting at-risk learners based on their performance-based behavior in e-learning environments. The proposed approach can identify and predict learners with difficulties and intervene autonomously to assist them in overcoming them. A novel algorithm is developed to forecast learners who are prone to struggle or drop out. We experimented in a learning environment at a higher education institution that used the proposed strategy to examine its effectiveness, and the findings supported the proposed approach's efficacy.