Design And Implementation Of A Potential-field-based Cluster Contraint Algorithm For Drawing Graphs

With the rise of SNS (Social Networking Services), neural networks, Deep Learningresearch become hot, these complex networks have been closely connected with our life.There is a lot of research on them, and in return a large number of results are received.Meanwhile, it is very important to know how to visualize them. Visualization can make iteasier for people to see the network structure, and extract a wealth of informationresources. Currently doing the visualization of the complex network with a specificstructure has been a very mature technology. However, when deal with the actualnetworks, the currently methods always get some drawbacks. Especially when thenetwork size is large, it is hardly to see the network structure. Moreover, as the existingpublicly available visualization software and frameworks tends to the theoretical studies(eg NetworkX, UCINET, etc.), or mainly for highly specialized applications (such asCytoscape for biological network visualization), and relatively good products aregenerally not open source and many of them are used in commercial operation (such asGoogle’s TouchGraph project), and the people in our country even can not get someproducts (eg VxInsight). So, developing a easy-to-use Complex Network Visualizationapplication based on existing theories would be very promising.This article provides an overview of the visualization of complex networks,describes their history and development status, and introduced several state-of-the-artclassic visualization methods. It then describes some typical network visualization toolsand development frameworks. Based on this, The article first summarize the discussionabout the common patterns and principles that most complex network visualizationalgorithms use, and give a analysis of the pros and cons of it, several examples of thevisualization results are given. Then according that the actual network has thecharacteristics of cliques and aggregation, the article presents a path-based multi-clusteraggregation constraint layout method, the purpose is to compress the densest part of the complex network, through a two-stage layout method, to get a clearer network structure,and the method can also be rendered in real-time visualization process. In this article themethod is described in detail. Finally, for the method this article provides, a basicframework of the complex network visualization is implemented through the C++language, and it presents the visualization process in real-time display using Direct2Dtechonology.This program strives to meet the dynamic visualization effects (display layoutchange status in real-time) in the GUI (Graphic User Interface). Some examples of therandom growth network with communities are presented, and the visualization resultsreflect the structures of the networks clearly.