IDGCN: A Proposed Knowledge Graph Embedding With Graph Convolution Network For Context-Aware Recommendation Systems

Context-aware recommender systems have been employed in heterogeneous domains and systems and have been implemented using a diverse range of approaches designed to realize personalization with targeted service provision in systems which are context aware. The diverse range of approaches used include knowledge graphs, which have gained traction for recommender systems driven by large complex datasets with user profiles, social network data, and logfile data are stored in directed heterogeneous knowledge graph(s) where nodes represent entities and edges correspond to relationships. However, the representation of entities generally fails to model the relationships that exist between entities. In this paper, we present a novel process to improve the efficacy of recommender systems employing knowledge graph embeddings with a convolutional network. In our proposed model, all user and item-based relationships are considered to enable the detection of the relationships that exist between them. To evaluate our proposed model, experimental testing has been implemented using “real-world” public datasets including a comparative analysis between the “state of the art” baseline approaches and the proposed method. Our reported experimental results indicate that our proposed model outperforms the alternative methods in user recommendations with enhanced targeted service provision and personalization.