Research On Missing Information Complementation Technology For Triple Based On Knowledge Graph

The advent of knowledge graph technology has provided the academic community with a more effective solution for organizing,managing,and leveraging large volumes of data on the Internet.A knowledge graph serves as a valuable information structure that connects knowledge,concepts,and entities.Presently,knowledge graph technology finds widespread application in various domains of artificial intelligence,including information retrieval,recommendation systems,intelligent question answering,natural language processing,and intelligent decisionmaking.However,one persistent challenge in knowledge graphs is their completeness.Many implicit relationships between entities often remain unexplored,which hampers the effectiveness of knowledge graphs in applications such as information retrieval and recommendation services.This incompleteness results in reduced accuracy and hinders the overall user experience with intelligent services.Consequently,it becomes imperative to address the data completion aspect of knowledge graphs,aiming to infer implicit relationships between entities and make the graphs more comprehensive.As a result,knowledge graph completion has emerged as a significant research focus in the field of artificial intelligence.This paper comprehensively surveys and introduces knowledge graph completion techniques,outlines the challenges they face,presents the current research status and progress both domestically and internationally,and explores future development directions.And the current methods of translation model,knowledge representation learning,and knowledge inference based on the knowledge graph are analyzed,and then the improvement algorithm model is proposed for the problems of the current graph embedding model,while the algorithm model is applied to the complementary prediction system.Finally,the algorithmic model is implemented in a supply chain management recommendation system.The main research of this paper are as follows:1)The problem is that the current commonly used knowledge graph-based translation models only consider a single triplet information in tasks such as graph construction is analyzed,ignoring the complex relationships among different triplet entities and relationships.A new entity and relation-based embedding model is proposed,which maps entities and relations of different triples in different Spaces and preserves their multidimensional vector representation;2)Describing the problems such as the huge workload and the tendency of missing information when training large datasets containing complex triadic relationships,a Trans E embedding model with distributed training combined with graph neural networks(GNN)is proposed.improved the training efficiency,dealt with the complex relationships existing between different entities,and validated the algorithm on the datasets FB15 K and WN18;3)A predictive system for knowledge graph completion,specifically targeting missing information regarding entities and relations,was designed and implemented.By training and testing on a dataset of triplets,the system is able to predict missing triplets and determine whether to add them to the knowledge graph,thus achieving knowledge graph completion.In combination with the completion prediction system,a recommendation architecture based on knowledge graph was constructed.The functionality and performance of this architecture were tested and analyzed.Finally,the architecture was applied to a recommendation system in the field of supply chain management.