| The development of information technology can effectively alleviate the professional problems of triage caused by the shortage of medical resources.Knowledge graph can effectively represent medical knowledge and its association,and the triage process is similar to the entity reasoning process of knowledge graph,that is,the reasoning of target entity is completed according to the given entity and specified relationship,so it is suitable for the research of triage.Rule-based entity reasoning is highly specialized in medical treatment,but it still has the following shortcomings in the process of triage,that is,the two problems of path effectiveness differentiation and large computational cost.Naive Bayes can effectively deal with entity reasoning by relying on the simple assumption of independent attribute conditions,Therefore,it can be used as a step of path effectiveness differentiation to improve the existing entity reasoning methods,and the topological properties of complex networks can also be used to analyze the association structure in the knowledge graph to reduce the computational cost of the algorithm.Therefore,this thesis applies their advantages to the entity reasoning process to improve the reasoning effect of triage.Based on the research of rule-based entity reasoning algorithm,this thesis makes corresponding improvements.The main research work includes:(1)A knowledge graph entity reasoning on Naive Bayes(KENB)algorithm is proposed.In view of the insufficient effectiveness of entity path selection in the current research,based on the construction of partial weighted knowledge graph based on Naive Bayes and the integration of subgraph feature extraction(SFE),a calculation method to measure the effectiveness of different subgraphs and a rule reasoning method for multiple entities are proposed,Solve the problem of low effectiveness of path selection of unauthorized knowledge graph,and reduce the number of random paths by limiting the related paths of entity nodes.The experimental results verify the rationality of the construction of weighted knowledge graph and the effectiveness of KENB solution path selection.(2)A knowledge graph entity reasoning on topological properties of complex networks(KETCN)is proposed.Aiming at the problem that KENB fails to effectively reduce the cost of path selection,on the basis of constructing a fully weighted knowledge graph using instance data,combining the topological properties in complex networks and the characteristics of the association structure of knowledge graph,an entity reasoning method based on purity and correlation is proposed to solve the computational cost problem caused by retaining the path traversal mode of SFE algorithm in KENB,At the same time,effective subgraphs are screened to reduce the interference of inefficient subgraphs.The experimental results verify the rationality of using topological properties and the effectiveness of KETCN to further improve path selection. |
