Implementasi Content-Based Filtering untuk Sistem Rekomendasi Kursus Online Berbasis Web

The rapid growth of online learning platforms has led to an increasing number of available courses, making it difficult for users to find courses that match their learning needs and interests. To address this issue, this study develops a course recommendation system based on Content-Based Filtering using the Term Frequency–Relevance Frequency (TF-RF) weighting method and Cosine Similarity. The dataset consists of 8,419 Information Technology courses collected from various online learning platforms and utilizes course attributes such as title, description, topic, and difficulty level. The system development process includes dataset processing and labeling, text preprocessing, vector representation using TF-RF, course filtering based on topic and difficulty level, and similarity calculation using Cosine Similarity to generate the Top-10 most relevant course recommendations according to user preferences. System evaluation was conducted using 10 simulated user preference scenarios and assessed with Accuracy, Precision@K, Recall@K, and F1@K metrics. The evaluation results show that the system achieved an average accuracy of 0.8177, Precision@10 of 0.8850, Recall@K of 0.2391, and F1@K of 0.3206. The high Precision value indicates that most of the recommended courses are relevant to user preferences. The results demonstrate that the combination of TF-RF and Cosine Similarity can be effectively applied to build a content-based course recommendation system capable of providing recommendations based on user-selected topics, skill levels, and keywords without requiring user interaction history.