Design And Implementation Of A Community Evolution Analysis System On Linked Data

The closely connected node forms community in the network.Community can reflect the structure feature of network which is useful in optimizing network’s structure,understanding the network functionality and finding the undelying pattern of network.Howerver,the community mining study regarding linked data is few.The multi-mode multi-dimensional structure of linked data brings obstacle to the community detection task in the linked data.In the area of complex network analysis,the community analysis in dynamic network is a hot topic.Usually,there exists abundant of temporal information in linked data,which lays foundation for the dynamically analysis of linked data.In this thesis,the linked data community evolution analysis system is presented based on the research method of complex network on dynamic network and the feature of linked data.First,through the extraction of the structural and unstructural temporal information from the linked data,the linked data temporal model is built.Then,by generalizing the idea of collaborative network,the ARC model is proposed in the thesis.Based on the ARC model,the method for constructing collaboration network for different time window is given.Besides,a distance measurement method of entity based on the tensor factorization of linked data is presented.Combined with the centrality-based community detection method,it is used to detect community on each slice.For the community evolutionary behavior analysis,a more rational community evolutionary event definition is proposed and used for event detection.Finally,the community detection and evolutionary event detection result is visually presented to user,which helps the further study of the community.The study of a certain type of entity’s community evolution in dynamic collaboration network can be implemented through this system,which brings a new idea and method in linked data mining and has both theoretical and practical value.