Core Concepts
Introducing Dual-Channel Multiplex Graph Neural Networks (DCMGNN) to address behavior patterns in recommender systems.
Abstract
Efficient recommender systems play a crucial role in capturing user and item attributes that mirror individual preferences. Existing techniques struggle with modeling behavior patterns formed by multiplex relations between users and items. DCMGNN introduces a novel framework to address these challenges, outperforming state-of-the-art methods. The model incorporates explicit behavior pattern representation learning and relation chain-aware encoder to capture the impact of various relations on target relations.
Stats
DCMGNN surpasses best baselines by 10.06% and 12.15% on average across all datasets.
Retail dataset: Users - 2,174, Items - 30,113, Interactions - 9.7 × 104.
Tmall dataset: Users - 15,449, Items - 11,953, Interactions - 1.2 × 106.
Yelp dataset: Users - 19,800, Items - 22,734, Interactions - 1.4 × 106.
Quotes
"Our DCMGNN model significantly outperforms existing recommendation methods."
"The proposed framework captures behavior patterns formed by multiplex relations between users and items."