This study discusses the use of an online learning recommendation system as a smart solution related to changing the face-to-face learning process to online. This study uses user-based collaborative filtering, item-based collaborative filtering, and hybrid collaborative filtering. This research was conducted in two stages using the KNN machine learning algorithm: (1) the three methods were tested to obtain student grade prediction results without adding contextual information; and (2) with the same method the same steps were carried out but with the addition of contextual information features as a feature addition. One of the alternatives carried out in this study is related to the possibility of predicting student grades. This study proves that the use of contextual information as an additional feature in the recommendation system has a significant effect on the accuracy of student score prediction results, which are used as the basis for providing recommendations using the rule base technique.