Research And Implemetation Of Graph Network Method Based On Deep Learning

In recent years,more and more attention has been paid to the research of graph network,and the effective practice of graph network method can often be seen in the latest research in various fields.This paper observes that these studies are divided into two categories:the first category uses graph networks as the research object,with the purpose of researching more effective methods on graph networks and enhancing the ability to interpret graph networks;the second category is using graph network technology to solve other problems.Further analysis concluded that the two categories are actually facing the problem of incomplete or missing graph structure,which hinders the application of graph network and the effective use of graph network methods.In view of the similarities and differences between the above two categories of tasks,this paper proposes an adaptive graph network technology,which designs models on specific knowledge graph completion tasks and urban grid supply and demand prediction tasks,and both are used in their respective tasks.It has been promoted,which verifies the versatility and effectiveness of the technology.Specifically include the following three achievements:(1)A knowledge graph complement model based on adaptive graph structure is proposed,which improves the neural network complement model based on embedding,and uses adaptive graph technology to model each dimension of entity and relationship embedding,and enhance the model The ability to express fine-grained associations.Experiments on link prediction tasks show that the improved model is better than the original model in all indicators.(2)A prediction model of urban grid area supply and demand based on spatial-time modeling is proposed,which uses adaptive graph technology to integrate time,space,and time-spatial graph network information to make prediction.The model performed well on the actual business data of the AutoNavi taxi platform and was proven to have better adaptability.(3)Designed and implemented a medical knowledge graph query system based on the completion technology,explored the algorithms and technical support required for the application of knowledge graphs,and verified the effectiveness of the proposed model in the practical application of knowledge graphs.