<![CDATA[Recommendation systems have become a crucial component in enhancing user experience on e-commerce platforms, including PT Renos Marketplace Indonesia. However, the existing recommendation system was still random and unable to adapt to user preferences. This study aims to develop a product recommendation backend system based on Collaborative Filtering using the Singular Value Decomposition (SVD) algorithm. The methodology employed is Research and Development (R&D) with an Agile Scrum approach, along with performance evaluation using RMSE, MSE, and MAE metrics. The system was developed using FastAPI, PostgreSQL, and Python, and its functionality was tested using black-box testing methods. The implementation results show an RMSE of 0.1886, MSE of 0.0366, and MAE of 0.1242, indicating excellent prediction accuracy. Additionally, a user perception survey involving 11 internal respondents showed an average satisfaction score above 80%, with the highest score of 90.91% for the statement indicating that the system increases purchase likelihood. These findings demonstrate that the SVD algorithm is effective in generating relevant and personalized recommendations. The study concludes that the developed backend recommendation system successfully improves the efficiency and relevance of product recommendations and opens opportunities for further development through the integration of more complex user behavior data.]]>
Copyrights © 2025
