A Hierarchical Knowledge and Interest Propagation Network for Recommender Systems

Chen, Qinghong; Tan, Huobin; Lin, Guanghui; Wang, Ze

doi:10.1109/icdmw51313.2020.00026

Cited by 3 publications

(9 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Compared to the state-of-the-art baseline model, the proposed model improved the AUC metrics by 3.36% and 2.49% on the Book-Crossing and Last.FM datasets, respectively, and the F 1 metrics by 6.73% and 3.18%, respectively. Specifically, compared to the model of literature [ 20 ] and the AUC and F 1 values of the model of literature [ 17 ] incorporating the attention mechanism, the model of literature [ 21 ] and this paper show significant improvements, indicating that the recommendation model incorporating the attention mechanism can more accurately learn the embedding representation of users and items during the dissemination process. The model of literature [ 16 , 17 , 20 , 21 ] and the model proposed in this paper all use propagation for personalized recommendation.…”

Section: Results Analysis and Discussionmentioning

confidence: 93%

“…Specifically, compared to the model of literature [ 20 ] and the AUC and F 1 values of the model of literature [ 17 ] incorporating the attention mechanism, the model of literature [ 21 ] and this paper show significant improvements, indicating that the recommendation model incorporating the attention mechanism can more accurately learn the embedding representation of users and items during the dissemination process. The model of literature [ 16 , 17 , 20 , 21 ] and the model proposed in this paper all use propagation for personalized recommendation. Cause of the model in literature [ 16 ] performs higher-order propagation representation learning unilaterally from the user's perspective.…”

Section: Results Analysis and Discussionmentioning

confidence: 93%

“…To further validate the performance metrics of the model, the experimental results of the proposed model are compared with other state-of-the-art models such as the Ripplenet model of literature [ 16 ], KGAT model of literature [ 17 ], KGCN model of literature [ 20 ], HKIPN model of literature [ 21 ], and PHGR model of literature [ 22 ].…”

Section: Results Analysis and Discussionmentioning

confidence: 99%

“…The advantage of the model in this paper is that it performs higher-order propagation representation learning from the perspective of user-project interactions on a collaborative knowledge graph. The models in literature [ 17 ] and literature [ 21 ] fuse user-item interaction information into the knowledge graph for obtaining higher-order information about users and items. The model in this paper builds on this by using user-item interaction information in interlayer propagation to further improve knowledge relevance in the higher-order propagation process.…”

Section: Results Analysis and Discussionmentioning

confidence: 99%

“…e models in literature [17] and literature [21] 0 fuse user-item interaction information into the knowledge graph for obtaining higher-order information about users and items. e model in this paper builds on this by using user-item interaction information in interlayer propagation to further improve knowledge relevance in the higher-order propagation process.…”

Section: Comparison Experimentsmentioning

confidence: 99%

See 4 more Smart Citations

A Multimodal Fusion Online Music Education System for Universities

Liu

Cao

Wang

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

In the context of Internet technology, the integration of information technology and education is a powerful supplement to the traditional teaching model of higher education. Online learning has become the new development direction of the education industry in the network era. To address the problems of serious difficulty in completing online teaching tasks, difficulty in monitoring teaching effects, and fragmentation of course resources in universities, a multimodal music knowledge graph is constructed. A personalized learning strategy based on users’ interest is proposed through the mining of online education data, and a music online education system has been developed on this basis. To improve the recommendation accuracy of the model, an embedding propagation knowledge graph recommendation method based on decay factors is proposed. The model considers the changes in the strength of user interest during the intra- and interlayer propagation of the knowledge graph interest map and focuses on higher-order user potential interest representations for enhancing the semantic relevance of multihop entities. The experimental results show that the proposed model brings a good prediction effect on several benchmark evaluation metrics and outperforms other comparative algorithms regarding recommendation accuracy.

show abstract

Section: Results Analysis and Discussionmentioning

confidence: 93%

Section: Results Analysis and Discussionmentioning

confidence: 93%

Section: Results Analysis and Discussionmentioning

confidence: 99%

Section: Results Analysis and Discussionmentioning

confidence: 99%

Section: Comparison Experimentsmentioning

confidence: 99%

See 3 more Smart Citations

A Multimodal Fusion Online Music Education System for Universities

Liu

Cao

Wang

2022

Computational Intelligence and Neuroscience

View full text Add to dashboard Cite

show abstract

KEMIM: Knowledge-Enhanced User Multi-Interest Modeling for Recommender Systems

Yang

Yue

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Researchers typically leverage side information, such as social networks or the knowledge graph, to overcome the sparsity and cold start problem in collaborative filtering. To tackle the limitations of existing user interest modeling, we propose a knowledge-enhanced user multi-interest modeling for recommender systems (KEMIM). First, we utilize the user-item historical interaction as the knowledge graph's head entity to create a user's explicit interests and leverage the relationship path to expand the user's potential interests through connections in the knowledge graph. Second, considering the diversity of a user's interests, we adopt an attention mechanism to learn the user's attention to each historical interaction and each potential interest. Third, we combine the user's attribute features with interests to solve the cold start problem effectively. With the knowledge graph's structural data, KEMIM could describe the features of users at a fine granularity and provide explainable recommendation results to users. In this study, we conduct an in-depth empirical evaluation across three open datasets for two different recommendation tasks: Click-Through rate (CTR) prediction and Top-K recommendation. The experimental findings demonstrate that KEMIM outperforms several state-of-the-art baselines.

show abstract