On Recommending Category: A Cascading Approach

This paper proposes a cascading category recommender (CCRec) model with a variational autoencoder (VAE) to perform category-level recommendations in e-commerce, which complements item-level recommendations.

Details

The paper discusses the importance of category-level recommendation in e-commerce, as it can enhance user experience and boost commercial success by allowing platforms to explore users' potential interests beyond just individual items. Existing works have mostly focused on item-level recommendations, but have started to incorporate category-level preference prediction to aid item-level recommendation. However, the authors argue that simply applying item-level models to category-level recommendation ignores key differences between the two tasks. To address this, the paper proposes a cascading category recommender (CCRec) model that uses a variational autoencoder (VAE) to encode item-level information and perform category-level recommendations. Experiments show the advantages of this approach over methods designed for item-level recommendations.