A Visual Link Analysis System On Linked Data

As the core data model of semantic web,linked data use Resource Description Framework(RDF)to form the relationship among data.After several decades of development,linked data has expanded quickly and its scale has become bigger and bigger.In order to be able to effectively carry out knowledge discovery from large-scale linked data,it is extremely necessary to do link analysis for linked data.Common link analysis tasks can help users to fully understand the structure of linked data,so as to provide guidance for further knowledge discovery.However,there still exists following problems to be solved:1)RDF model is a Multi-dimensional Multi-mode model,so its complexity dose not suitable for large-scale linked data analysis.There is still in need of an effective linked data analysis model;2)Some linked data analysis methods,like traditional graph clustering analysis,cannot effectively process large-scale linked data;3)The existing link analysis methods are designed for comparatively small-scale linked dataset,and are in lack of an automatic,customized analysis platform.Besides,the analysis result cannot be easily visualized,which makes it hard for users to understand the structure of linked data through the analysis result.Based on above issues,we design and implement a visual system about link analysis.The contribution of this thesis is on the following aspects.(1)Propose a model of Object Graph derived from RDF graph,which can solve the problem from Multi-dimensional Multi-mode model.And then the notion of Virtual Document is used to measure linguistic closeness between objects.(2)Propose a scalable approach of centrality-based clustering,which include distance-maximization strategy for centroid selection and LPA-based Clustering,and have verified that it can perform effective clustering analysis for large-scale linked data.(3)Design and implement a visual system about link analysis based on B/S architecture that is named LIV.The system can analyze,transform and store the structure of the link data,and automatically complete link analysis tasks so as to provide users friendly visualization and interactive interface.Besides,the system can help users understand the structure of linked data with good theoretical and engineering value.