ABOUT

PEOPLE

ADMISSION

EDUCATION

ARCHIVE

BOARD

  • ARCHIVE

MV CIMF Festival

Hypergraph temporal multi-behavior recommendation

Jooweon Choi , JuneHyoung Kwon , Yeonghwa Kim, YoungBin Kim Intelligent Information Processing Lab

Modern e-commerce users perform diverse actions—viewing, favoriting, adding to cart, and purchasing. However, traditional recommendation systems treat these behaviors uniformly, neglecting their temporal evolution and high-order dependencies, which leads to lower accuracy and severe data sparsity. To overcome the limitations of traditional recommendation systems—particularly accuracy degradation and data sparsity—by jointly modeling time-varying behavioral dynamics and cross-behavior relations, enabling the system to understand when and how user intentions emerge rather than merely what to recommend.