The Research And Implementation Of Network Relational Data Visualization

Network visualization is one of the important technological means of network analysis. Visualization techniques can help people to better understand the network's inner construction and the connection along vertexes, which provides a service to find the law behind the network.After entered the age of big data, the network scale has become lager, and its structure is becoming more and more complex. As for this large-scale network, it is difficult for researchers to get valuable information form the direct visualization. Therefore, it is necessary to visualize the network after the network hierarchical processing.Layer division of the large scale network was studied in detail in this paper. Louvain detect community algorithm was selected as the fundamental algorithm to create hierarchical community structure by comparing the existing hierarchical partitioning strategy. However,Louvain algorithm is not suitable for processing the large-scale network data and has limited support for parallel computers, because this algorithm itself greedy iterative process and is not friendly to the distributed framework. We improved the Louvain algorithm from serial algorithms to parallel algorithms by adding single degree pretreatment,and dividing independent sets, and we proposed a community detection algorithm based on Spark distributed computing framework in this paper.Compared with the Louvain algorithm, the distributed algorithm proposed in this paper can process the large-scale network data, and detect community structure quickly.After the network is spilt into hierarchical community structure,communities can be visualized as vertexes. Because the number of the communities is far smaller than the number of the vertexes, this visualization has good processing capacity for large scale network. At the base of force-directed layout FR algorithm, the researcher of this paper limited the scopes of repulsion and gravitation, which made FR algorithm more fit to display the network. After these two algorithms were implemented, a visualized network system based on Spark framework was built in the paper to display network for subsequent analysis.