Yuri Ariyanto, Triyanna Widiyaningtyas, Ilham Saifudin, Wiwik Suharso, Wahyu Caesarendra, Adinda Salsa Leviona
This study presents a demographic-enhanced collaborative filtering approach for improving the accuracy and diversity of movie recommendation systems, evaluated on the MovieLens 100K and MovieLens 1M datasets. The proposed method integrates K-Nearest Neighbors (KNN) clustering for user demographic profiling with matrix factorization-based collaborative filtering using Singular Value Decomposition (SVD). The system is evaluated against established baselines (SVD, User-Based CF, Neural CF) across four key performance metrics: precision, recall, F1-score, and diversity. Experimental results demonstrate that the proposed method consistently outperforms baseline models, with significant improvements in accuracy metrics when scaling from MovieLens 100K to 1M. Precision increased by 3.91%, recall by 20.99%, and F1-score by 15.12%. Notably, the recall improvement indicates an enhanced capability to capture diverse user preferences. However, the diversity score remained constant at 0.6666 across both datasets, suggesting that algorithmic design, rather than data volume, predominantly influences recommendation variety. Comparative analysis reveals that while the proposed method excels in accuracy, it yields slightly lower diversity than User-Based CF, highlighting a potential trade-off. These findings confirm that incorporating demographic data enhances collaborative filtering effectiveness, particularly for addressing cold-start issues and capturing broader user preferences. This research contributes to more accurate and personalized recommendation systems by leveraging demographic insights within a scalable collaborative filtering framework. © 2025 IEEE.
Universitas Negeri Malang, Department of Electrical Engineering and Informatics, Malang, Indonesia; Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia; Universitas Muhammadiyah Jember, Informatics Engineering, Jember, Indonesia; Curtin University Malaysia, Faculty of Engineering and Science, Department of Mechanical and Mechatronics Engineering, Lot 13149, Block 5, Kuala Baram Land District, CDT 250,Sarawak, Miri, 98009, Malaysia
