Triyanna Widiyaningtyas, Ilham Saifudin, Ilham Ari Elbaith Zaeni, Moh. Zulfiqar Naufal Maulana, Wahyu Caesarendra
In the current era, ranking-based collaborative filtering approaches have become the backbone in the development of recommendation systems. One of the widely applied methods is the memory-based method that focuses on item ranking. Recently, a new algorithm has been introduced that combines rating value prediction from the Singular Value Decomposition (SVD) method with similarity between users. However, this approach still faces limitations because it only relies on rating weights, so its accuracy is not optimal. To answer this challenge, this research offers a new algorithm with a ranking-based collaborative filtering approach that integrates matrix factorization through SVD with a ranking technique that combines rating values that have been normalized based on the number of users, as well as Gower similarity scores that are also normalized based on user representation. This algorithm is called SVD-GSetRank. Based on experimental results on MovieLens-100K, MovieLens-1M, Book-Crossing, Ciao, Epinions, and Flixster datasets, this approach shows superior performance in Top-N accuracy, especially in NDCG, Precision, Recall, Hit Rate, and MRR metrics, which are the main indicators in memory-based ranking recommendation systems. In addition, the algorithm also shows efficiency in processing time. © The Author(s) 2025.
Department of Electrical Engineering and Informatics, Faculty of Engineering, Universitas Negeri Malang, Malang, 65145, Indonesia; Department of Informatics Engineering, Faculty of Engineering, Universitas Muhammadiyah Jember, Jember Regency, 68121, Indonesia; Department of Mechanical and Mechatronics Engineering, Faculty of Engineering and Science, Curtin University, Miri, Malaysia
