 The proposed knowledge-enhanced Hierarchical Graph Capsule Network, CANC, is a novel approach to addressing the problem of cold start and sparsity in recommender systems. It utilizes knowledge graphs to capture the relationships between items and entities, and then uses a Graph Capsule Network to learn the representation of these relationships. This allows it to better understand the context of each user's preferences, resulting in improved recommendations. This article was authored by Fuku and Chen, Guaixing Yin, Yusun Dong, and others.