Professor U Kang’s Research Team at Seoul National University Develops Product Recommendation Technology to Maximize Profits in Online Shopping Stores
- Capable of boosting sales and clearing inventory by recommending a diverse selection without excluding items
- Research to be presented at the top-tier data mining conference, WSDM 2025


▲ (From left) Professor U Kang, Interdisciplinary Program in Artificial Intelligence/Department of Computer Science and Engineering, and Jongjin Kim, Ph.D. candidate, Department of Computer Science and Engineering

Seoul National University College of Engineering announced that Professor U Kang’s research team from the Interdisciplinary Program in Artificial Intelligence and the Department of Computer Science and Engineering has developed a personalized sequential recommendation technology incorporating diversified recommendation.

Diversified recommendations are systems designed to ensure that items listed on e-commerce platforms are recommended to users without excluding certain products, which is essential to maximizing platform revenue. This concept has drawn attention in recent studies focusing on e-commerce optimization.

Previous studies have approached Diversified Recommendation by collecting information about each user's product preferences and then recommending items that other users don't like among those with similar preferences. However, this approach cannot take into account information about what users will prefer in the future, so it is difficult to guarantee the diversity of product recommendations in the future.

To address this, Professor Kang’s team introduced SAPID (Sequentially Diversified Recommendation via Popularity Debiasing and Item Distribution), a technology that provides diversified recommendations by accounting for the order of user interactions. SAPID operates by predicting future demand for each product based on users’ past purchase data, recommending products with potential for increased diversity. The researchers trained the SAPID model on a training dataset to recommend items by referring to the frequency of each product's appearance on the platform, without skewing toward popular products. This process allows SAPID to increase diversity by prioritizing products that have not yet been recommended to platform users or are less popular.

This technology has the potential to maximize expected revenue for e-commerce platforms requiring sequential product recommendations. Adjusting the display order of products on a shopping mall’s main page through SAPID could increase overall sales while reducing inventory. Professor Kang noted, “SAPID enhances both the accuracy and diversity of product recommendations, delivering high academic and practical value. It is expected to make a significant contribution to increasing sales and clearing inventory for online shopping malls and content providers.”

Funded by the Junghun Foundation, this research will be presented in March 2025 at the WSDM (Web Search and Data Mining) conference, a leading academic conference in data mining and machine learning.

In addition, Jongjin Kim, a Ph.D. candidate in SNU Department of Computer Science and Engineering and the study’s first author, is currently conducting research on integrating diversity in sequential bundle recommendation processes.


[Reference Materials]
"Sequentially Diversified and Accurate Recommendations in Chronological Order for a Series of Users", WSDM 2025
https://kdd2024.kdd.org/wp-content/uploads/2024/07/paper_22.pdf

[Contact Information]
Professor U Kang, Interdisciplinary Program in Artificial Intelligence/Department of Computer Science and Engineering, Seoul National University / +82-2-880-7254 / ukang@snu.ac.kr