<![CDATA[This study aims to analyze the performance of a recommendation system based on collaborative filtering using the cosine similarity method in a digital learning platform. The main issue addressed is the low effectiveness of users in discovering relevant learning materials due to the large number of available courses, modules, videos, and learning resources. This study employed a quantitative approach with a computational experimental design. The dataset was organized into a user-item interaction matrix consisting of 1,200 users, 320 learning items, and 18,742 valid interactions. The data were divided into training and testing sets using an 80:20 ratio. The evaluated models included user-based collaborative filtering, item-based collaborative filtering, and a popularity-based baseline model. Model performance was assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Precision, Recall, and F1-score. The results indicate that item-based collaborative filtering using cosine similarity achieved the best performance, with an MAE of 0.612 and an RMSE of 0.824. In the Top-10 recommendation evaluation, the model obtained a Precision of 0.364, a Recall of 0.291, and an F1-score of 0.323. These findings suggest that similarity among learning items is more stable than similarity among users. This study concludes that cosine similarity is a suitable baseline method for developing digital learning recommendation systems. However, further improvements are required to address challenges related to data sparsity, the cold-start problem, and evolving user preferences.]]>
