Sistem Rekomendasi Produk Menggunakan Collaborative Filtering pada Marketplace Digital

Abstract

The rapid growth of digital technology and marketplace transactions highlights the need for effective product recommendation systems to help users find relevant items based on their preferences. This study aims to develop a product recommendation system using the Collaborative Filtering method to enhance personalization and user experience. The method is selected for its ability to generate recommendations based on user behavior patterns and preference similarities without relying on detailed product attributes. The research uses user transaction data, including purchase history, product ratings, and search activity, and follows stages of data collection, preprocessing, user–item matrix construction, similarity calculation using cosine similarity, and recommendation prediction. The system is implemented using Python and MySQL to support large-scale data processing. Performance is evaluated using accuracy, precision, recall, and Mean Absolute Error. The results show that Collaborative Filtering produces relevant and personalized recommendations with satisfactory accuracy, improves search efficiency, and increases the likelihood of purchase, while outperforming conventional popularity-based methods, and offering contributions to data mining and artificial intelligence development in digital marketplace systems.