search for


Comparative study of recommender systems based on deep learning algorithms
Journal of the Korean Data & Information Science Society 2024;35:1-13
Published online January 31, 2024;
© 2024 Korean Data and Information Science Society.

Hak Rim Lee1 · Cheolyong Park2

12Department of Statistics, Keimyung University
Correspondence to: This article is extracted from Hak Rim Lee’s master thesis. This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (NRF-2019R1F1A1058723).
1 Master’s graduate, Department of Statistics, Keimyung University, Daegu 42601, Korea.
2 Corresponding author: Professor, Department of Statistics, Keimyung University, Daegu 42601, Korea. E-mail:
Received November 25, 2023; Revised December 15, 2023; Accepted December 15, 2023.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
A recommender system is a system that selects and presents users with relevant information or products based on their past behavior data or other relevant data. There are many types of recommendation systems, and in this study, we tried to find the best model for movie recommendation among five popular collaborative filtering models based on traditional and deep learning techniques. The IMDb data was used for analysis, and the training data, validation data, and test data were partitioned using random and time-dependent partition methods, respectively. Performance comparison was conducted using rating-based evaluation and ranking-based evaluation methods. The BE (baseline estimate) model was better than the SVD (singular value decomposition) model in the rating-based evaluation, while the BiVAE (bilateral variational autoencoder) model was best in ranking-based evaluation. Additionally, the random partition method showed superior performance compared to the time-dependent partition method.
Keywords : Collaborative filtering, deep learning, movie recommender system.