Research On Network Visualization Layout And Interaction Techniques Based On Context

With the continuous development of network science,network analysis is widely used in various fields to help analyse entity association,anomaly exploration,propagation patterns,prediction,etc.in complex systems.Network visualization is an efficient and intuitive approach to network analysis.The association modes between entities in complex systems are closely related to attributes,so visual analysis of multivariate networks(MVNs)composed of entity objects and their attributes presents more challenges than networks that only consider topological structures or single attributes.Visual analysis of multivariate networks faces the following challenges in network layout and interaction analysis: How to balance the presentation of network topology and attributes in layout.In the aspect of interaction analysis,how to cope with the increasing network size and provide guided interaction to help users to quickly locate the target object.In view of the above concerns,this paper investigates both the layout and interaction of multi-dimensional network visualisation and proposes the multi-dimensional spatial layout method and the context-based progressive interaction scheme.Based on the layout method and interaction scheme,a visual analysis system of multiple networks is designed and implemented,which can provide multiple granularity topological contexts of global network,community network and local network simultaneously for comparison and correlation analysis.The main work of this paper includes:(1)In terms of layout,given the limitations of single-dimensional network layout and the complexity of multi-attribute encoding,we propose a multi-dimensional spatial network topology layout method which combination of 3-dimensional,2.5-dimensional and2-dimensional topological layout schemes.The traditional force-oriented layout is improved to achieve a 3-dimensional force-oriented layout based on community structure(Louvain-3D FDA),which is used to present the overall structure of multi-dimensional network.Edge binding algorithm is introduced to optimize the 2.5-dimensional layout,and multiple attributes were represented by topology layers.The traditional 2-dimensional layout is used to present the local topology of multiple networks.A multi-dimensional space layout transformation method is proposed to maintain topological contextual consistency in the 3 dimensional spaces.Based on this layout method,three datasets are selected for layout experiments and visual analysis.(2)In the aspect of interaction,a progressive interaction scheme based on contextual contexts is proposed to address the problem that traditional network visualisation and analysis tools lack guidance for interactive exploration as the scale of the network continues to expand.The interaction scheme is based on the principle of top-down interaction,guiding users from global to local network analysis,comparing and correlating multiple topological contexts with multi-grid topological contextual analysis contexts during the analysis process.(3)Based on the above layout method and interaction scheme,a multivariate network visual analysis system Multi-Net Vis is designed and implemented.The system is based on the multi-dimensional spatial network topology layout method to design the topology analysis module,and based on a progressive interaction scheme to design the interaction module.In addition,the important features of network attributes are incorporated to design the attribute analysis module.In this paper,two data sets are selected to demonstrate the application scenarios of the system,and comparative evaluation experiments are designed to evaluate the system from multiple perspectives.