<![CDATA[The development of web-based marketplaces has increased the number of available products, making it difficult for users to find products that match their needs and preferences. This condition creates information overload, which can reduce the effectiveness of product searching in marketplaces. This study aims to implement the User-Based Collaborative Filtering method with Cosine Similarity in a product recommendation system for a web-based plastic bottle marketplace. The research used a quantitative method with an implementation approach using real rating data from the marketplace consisting of 108 ratings, 36 users, and 15 products with a sparsity level of 80%. The research stages included building a user-product matrix, calculating cosine similarity, selecting K=5 neighbors, and predicting ratings using weighted average. The results showed that the system was able to generate product recommendations based on user preference similarities with the highest similarity value of 0.9939 and the highest predicted rating of 5.0000. The study also found that high sparsity caused many user pairs to have only a few co-rated items, resulting in trivial similarity values of 1.0000. Therefore, the User-Based Collaborative Filtering method with Cosine Similarity can be implemented in a web-based plastic bottle marketplace to support personalized product recommendations.]]>
Copyrights © 2026
