A TransE-based model for entity and relationship prediction in knowledge graphs

Knowledge graphs continue to face challenges related to data sparsity, even as their scale rapidly expands. Existing methods often struggle with complex relationships and dynamic updates within these graphs. In this paper, we conduct an in-depth study of entity and relationship prediction for knowledge graphs using the TransE model. TransE plays a significant role in entity and relationship prediction tasks due to its simplicity and effectiveness; its core approach involves representing entities and relationships through translations in vector space. To validate this method, we conducted experiments on the FB15K dataset, where TransE demonstrated strong performance. For relationship prediction, TransE achieved a Mean Rank of 81.9809 and a Hits@10 score of 0.7876. For entity prediction, the model attained a Hits@10 of 0.3107 and a Mean Rank of 257.1265. These results confirm TransE’s effectiveness in knowledge graph prediction tasks. Future research will focus on enhancing the model’s generalization and adaptability.