Triyanna Widiyaningtyas, Heni Vidia Sari, Moh. Muzayyin Amrulloh, Ilham Saifudin, Dzannun Fansyiari Austi, Wahyu Caesarendra
Model-based Collaborative Filtering (Model-based CF) is one of the approaches in recommendation systems that uses mathematical or statistical models to predict user preferences for certain items based on interaction patterns between users and items. One of the techniques often used in Model-based CF is Matrix Factorization (MF). This study aims to apply Model-based CF using MF techniques in a music recommendation system. The MF algorithms used are Singular Value Decomposition (SVD) and Alternating Least Squares (ALS). The performance testing of both algorithms was carried out on the LFM-2b dataset. The evaluation metric uses Normalized Discounted Cumulative Gain (NDCG), Precision, and Mean Average Precision (MAP). The results of this study indicate that ModCF with ALS performs better than SVD with an NDCG value of 0.06519, a Precision value of 0.0843, and a MAP value of 0.03715. This shows that the architecture of the ModCF and the use of ALS as a solver provide performance advantages, especially in the context of music recommendation order and relevance. © 2025 IEEE.
Universitas Negeri Malang, Department of Electrical Engineering and Informatics, Malang, Indonesia; Curtin University Malaysia, Department of Mechanical and Mechatronics Engineering, Sarawak, Malaysia
