Automatic Generalization Methods of Cyberspace Point Cluster Features Considering Characteristics

  Objectives  The visualization of cyberspace information is of great significance to reveal the rules and promote the cognition of cyberspace. Visualizing the nodes and their topological relationships in cyberspace without preprocessing often result in a large number of point coincidences and line crossings. Although the visualization result can be optimized by the node layout algorithms, the divided edge bundling technologies, the backbone extraction technologies and the methods of multi-scale representation of network routing, the spatial characteristic information of these nodes as cyberspace point cluster features cannot be maintained on a micro level of network structure. Therefore, by analyzing and quantifying various types of characteristic information of point cluster features in cyberspace from four aspects of statistical information, metric information, topology information and thematic information, two generalization methods are proposed.

  Methods  (1) An aggregation method based on hierarchical clustering. First, organize the point cluster features into a hierarchical tree of which every final node community is in an appropriate and well-balanced size through multi-layer K-means clustering according to their topology information. Then aggregate these features by their community.(2)A selection method based on node importance measuring. After measuring the comprehensive importance of the each feature from the aspects of maintaining the structural stability and its service coverage area, select the more important ones and delete the relatively unimportant considering the whole network topology.

  Results  A series of generalization experiments at different spatial scales were performed on the point cluster features incyberspace of an area in a central part of China through these two methods of aggregation and selection. The result shows that these methods can realize the automatic generalization of the point cluster features in cyberspace at different spatial scales while maintaining the spatial characteristics of these features.

  Conclusions  The purpose is to provide an effective data processing method of quantitatively expressing the characteristics of cyberspace and accelerating the process of optimizing visualization results.