Research On Visual Analysis Method For Domain Knowledge Graph

Various domains have accumulated abundant data,but due to the lack of reasonable data management methods,there are still tow major challenges in the use of domain data.The first is the unified modeling of data with different structures,which integrates the decentralized data and provides a strong foundation for the effective utilization of data;The second is to provide efficient data analysis and knowledge acquisition methods for specific application scenarios to improve the efficiency of researchers.In view of the above problems,this paper studies the construction of domain knowledge graph and visual analysis method.The main research contents and achievements of this paper are as follows:(1)In specific application scenarios,due to the flat structure of the knowledge graph,the relationships of the knowledge graph will be intertwined,making it difficult to effectively extract knowledge and increase the difficulty of analysis.In order to better correspond to the application scenario,this paper proposes a method of constructing topic knowledge graph.Based on the constructed domain knowledge graph,combined with the knowledge system required by the application scenario,the entity is mapped into low-dimensional vectors through the proposed entropy2 vec graph embedding algorithm,and then the graph is divided into different sub graphs through similarity calculation.Finally,the improved Sankey graph is used for visual presentation.(2)In the visualization of linked data,the semantics of entity relationship is an important information to understand data.This paper proposes a relationship oriented visual interaction method for linked data.Firstly,by studying the relational semantics and different layout methods in linked data,the mapping between relational semantics and layout methods is established.Then,the user preferences are collected and analyzed according to the established mapping.Finally,based on the user preference and the mapping,the mixed layout method is used to recommend the layout of locally dense subgraphs.(3)Combined with the above work,this paper focuses on the formulation design of polymer bonded explosives,and develops a visual analysis system for the knowledge graph of explosive formulation design,which solves the problem that it is difficult to obtain and use the formulation design principle knowledge,component correlation knowledge and preparation process knowledge in the process of explosive design.The system designs visual analysis methods for knowledge graphs of different topics,so that the knowledge of explosive formula aided design can be used quickly.The proposed method provides a new means for the convenient acquisition and use of knowledge in explosive formula design,and provides an intuitive and easy-to-use auxiliary tool for researchers in explosive formula design.