부산대학교 도서관

Currently, research models that effectively predict cross-selling products while utilizing multimodal data sources are limited, and similarly, models focusing on multimodal recommendation do not adequately address cross-selling. To address this gap, our study introduces the model HHMC: A Heterogeneous x Homogeneous Graph-based Network for Multimodal Cross-Selling Recommendation. This innovative approach leverages historical order's data and diverse multimodal data to recommend cross-selling products. The architecture of model HHMC is thoughtfully designed to explore potential relationships between user-item and item-item interactions while also improving the efficiency of item feature representation through the enrichment of multimodal data sources. Due to the scarcity of published datasets for the cross-selling recommendation problem, we utilized the well-known Instacart dataset to define and explore empirical directions for addressing this challenge. Experimental results demonstrate that HHMC surpasses widely used deep learning-based techniques, highlighting its potential to effectively address multimodal cross-selling recommendation problems.