KAIST AI Model Wins Top Paper Award at IEEE ICDM, First for Korean University in 23 Years

A Korean research team has developed an artificial intelligence model that predicts how individual characteristics such as age, roles, and interests determine collective behavior, winning the best paper award at the world's most prestigious data mining conference.

KAIST announced Wednesday that Professor Shin Ki-jung's team at the Kim Jaechul Graduate School of AI developed an AI model called NoAH (Node Attribute-based Hypergraph Generator) that precisely predicts and reflects how people behave when they gather in groups.

NoAH analyzes both "relationships between individuals" and "individual characteristics" simultaneously. While previous AI models focused on analyzing either personal characteristics or relationships, NoAH observes and analyzes both at the same time. This enables the model to identify how group interactions involving multiple participants—such as online communities, research collaborations, and group chats—are structured and how individual characteristics influence these formations.

NoAH explains and replicates what group behaviors emerge when people's characteristics come together. For example, the model identifies what interests people have and what roles they play, then analyzes how this information actually converges to create group behavior and reproduces it accordingly.

The model has demonstrated significantly more realistic reproduction of various real-world collective behaviors compared to existing models, including purchase combinations in e-commerce, spread patterns in online discussions, and co-authorship networks among researchers.

The research team achieved the distinction of winning the best paper award at IEEE ICDM, a leading global data mining conference organized by the Institute of Electrical and Electronics Engineers. The award is given to only one paper among 785 submissions worldwide. This marks the first time a Korean university research team has received this honor since Professor Kim Yong-dae of Seoul National University's Department of Statistics won in 2002, a gap of 23 years.

"This research opens a new AI paradigm that enables three-dimensional understanding of complex interactions by considering not only group structure but also individual characteristics," Professor Shin said. "Analysis of online communities, messengers, and social networks will become much more precise."

The research was conducted by a team comprising Professor Shin Ki-jung and KAIST Kim Jaechul Graduate School of AI master's students Jeon Jae-wan and Yoon Seok-beom, along with doctoral students Choi Min-young and Lee Geon. The findings were presented at IEEE ICDM on November 18.