The Research And Implementation Of Knowledge Model And Knowledge Graph Inference

With the rapid development of the IoT in the information age,research and applications related to it are in full swing.Obj ects in the IoT have their own corresponding abilities,and there is a certain relation between these abilities.In order to understand the abilities of the IoT more comprehensively and deeply,this paper uses knowledge graph technology that has strong capabilities in semantic understanding.This paper first proposes a method based on the characteristics of the IoT to construct a knowledge model based on the capabilities of obj ects in the Io,T.In the logical structure of the knowledge graph,a joint method based on top-down and bottom-up is used.And the data layer is constructed by using web crawler technology.Then this paper studies and implements a knowledge inference algorithm based on the potential probabilities of the negative triples of the relational path.According to the known problems of the proposed algorithm,two possible solutions are proposed:linear smoothing potential probability and adaptive transformation potential probability.This paper takes an experiment on four public datasets:FB15k,FB15k-237,WN18,WN18RR and the knowledge graph dataset.The experiment result indicates that introducing the potential probability of the negative triple to the subsequent calculation,the performance of the knowledge inference algorithm on the two evaluation criteria of Mean Rank and Hits@10 of the link prediction task can be greatly improved.Finally,this paper designs and implements a prototype system for knowledge management and display including user management,knowledge map management,knowledge graph visualization,data editing,and knowledge inference.The system uses Neo4j graphics database as the underlying data storage tool.The Bootstrap and the Spring Boot are used to complete the system front and back end.The d3.js visual JavaScript library is used to complete the visualization of the system.