Citation

Horani, Amany Sharaf Al and Goh, Chien Le and Horani, Sharaf Sami Al (2026) A Workbench to Study Movie Recommender Systems. In: 2026 International Conference on Smart Multidomain Integrated Learning Environments, ICSMILE 2026, 30 March 2026 - 31 March 2026, Irbid.

Text 2.pdf - Published Version Restricted to Repository staff only Download (435kB)

Abstract

Recommender systems are widely used across digital platforms, yet practical tools for comparing different recommendation techniques remain limited. This paper presents a workbench that supports experimentation with collaborative filtering, content-based filtering, and hybrid recommendation strategies. The system implements SVD++ for collaborative filtering, TF-IDF with cosine similarity for content-based filtering, and two hybrid approaches: weighted and switching methods. The workbench is built using a Python backend and D3.js frontend, allowing for interactive visualization and side-by-side evaluation. Experiments on MovieLens datasets (1M, 10M, 25M ratings) demonstrate that collaborative filtering achieves the lowest RMSE at 0.9101, outperforming content-based (RMSE=1.0648) and hybrid approaches. The hybrid weighted method with α=0.7 balances accuracy (RMSE=0.9410) and diversity, while contentbased filtering offers 55x faster training (7.09s vs 393.63s). The workbench provides a modular architecture supporting reproducible experiments with configurable hyperparameters, serving as both an experimental platform and educational tool for understanding recommender system behavior across varying dataset sizes and algorithm configurations.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Recommender systems
Subjects:	Z Bibliography. Library Science. Information Resources > ZA3038-5190 Information resources (General)
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	30 Jun 2026 02:25
Last Modified:	30 Jun 2026 02:25
URII:	http://shdl.mmu.edu.my/id/eprint/16114

Downloads

Edit (login required)