Research On Knowledge Graph Construction And Auxiliary Recommendation Algorithm

Radar and communications are relatively independent but closely related.It is a hot issue in the radar field to combine radar and communication functions in one device integrate detection and real-time communication.The primary premise of radar communication integration lies in the fusion of knowledge in two fields.However,the effective tools to integrate the basic concepts of radar and communications are rare.Knowledge graph stores the entities and the relationship between entities in the form of triple and the unstructured knowledge can be represented by graph structures.This thesis constructs knowledge graph based on the basic concept of radar and communications,which promotes the integration of the two fields.Because knowledge graph can store rich structured data,the research integrates knowledge graph of radar and communications into the recommendation system to improve the recommendation performance.In this case,the relevant knowledge of radar and communications can be directly recommended by the knowledge graph for researchers in this field.To further verify the recommendation algorithm,the travel knowledge graph is constructed to recommend travel plans for people during the epidemic in view of the current hot issue about COVID-19.The main work of this thesis is given:1.Two methods of constructing a knowledge graph of radar and communications are studied top down and bottom up.For the top-down knowledge graph,it is first constructed by building hierarchical relationships between classes in the field of radar and communications using the Protégé tool.Then the objects and data properties are built for each class.Finally,a knowledge model is created by filling each class with entities and visualizing them on Onto Graf.For the bottom-up knowledge graph,it is first constructed by using the Stanford Core NLP tool to recognize named entities,to extract relationship and to extract other techniques to obtain the triples.Then the artificial knowledge fusion is implemented.Finally we use the graph database Neo4 j for knowledge storage and visualization.2.Two recommendation algorithms are studied based on the structure and feature of knowledge graph.The recommendation algorithm based on knowledge graph's structure(KGS-RS)uses the knowledge graph's structure which is similar to water ripple.This algorithm regards the user's history as the center of ripple,and spreads preferences outward along the connection of knowledge graph to explore the potential interests of the user.The recommendation algorithm based on the knowledge graph's features(KGF-RS)combines the entities and projects features by mapping them to the low-dimensional vector space.This algorithm combines the knowledge graph module with the recommended module through the cross-compression module.In this thesis,Book-Crossing and Bing News are selected as data sets and knowledge graph of radar and communications are used as auxiliary information.Simulation results show that these two recommendation algorithms have improved the performance of AUC,accuracy,precision and recall compared with the benchmark algorithm,and verify the integrity and accuracy of knowledge graph.3.Travel recommendation based on a travel knowledge graph in COVID-19 background is studied.First,the travel knowledge graph is constructed.Then the predicted infection population curve of the provinces and cities are obtained by comparing the analysis models.Finally,the appropriate travel plans for users are recommended by combining the predicted number of infected people with the constructed travel knowledge graph.