Research On Large Scale Network Data Visualization Technology

With the advent of the big data era,the scale of data in various fields has exploded.Since most of the relational data in the real world can be modeled in the form of a network,so,to effectively mine the information hidden in the network data and obtain more potential value from the data,Large-Scale network data visualization technology Research has received widespread attention from all walks of life.The network layout algorithm based on the Force-Directed model occupies a pivotal position in network visualization because of its intuitive layout results and easy analysis.However,the traditional network layout visualization technology cannot easily meet the needs of researchers for analyzing Large-Scale network data.For this reason,this paper has made in-depth research on Large-Scale network data visualization technology.This article first introduces the basic theories and related technologies of network data visualization and summarizes the challenges faced by current network visualization technologies in processing Large-Scale network data based on the research status at home and abroad.In response to these important challenges,this paper proposes a PageRank-based network layout algorithm to improve the quality and efficiency of Large-Scale network layouts.To reduce the computation time of LargeScale network layout algorithms more effectively,a heterogeneous parallel computing framework is proposed in this paper,and the effectiveness of this framework is proved by experiments.Finally,based on the above two results,this paper designs a related visual analysis system for efficient exploration of Large-Scale network data.My main contributions are in the following three areas:(1)PageRank-based network layout algorithm.This paper proposes a layout algorithm based on PageRank,which belongs to a Force-Directed model.The algorithm introduces PageRank to improve the calculation of nodes' gravity and repulsion to improve the quality of the layout;then introduces the Closeness Centrality to estimate the position of the nodes in the initial layout,so that the algorithm obtains a high-quality initial layout before starting the calculation of the force At the same time,an adaptive step length based on PageRank is proposed to balance the efficiency and quality of the layout algorithm.Finally,experiments prove that the algorithm can get a high-quality layout result when processing Large-Scale network data.(2)Heterogeneous parallel computing framework.When processing Large-Scale network data of hundreds of thousands or even millions of nodes,most optimization algorithms cannot obtain layout results in a short time.Because the Force-Directed model network layout algorithms,such as PageRank-based network layout algorithms,require a large number of simple iterative calculations during the layout phase,which is more suitable for GPU parallel computing technology.this paper proposes a CPU + GPU heterogeneous parallel computing Framework,control-intensive logic is executed on the CPU,computation-intensive logic is executed on the GPU,and communication is performed through the PCI-E bus.According to the experimental evaluation,it can be concluded that the heterogeneous parallel computing framework proposed in this paper can effectively reduce the layout calculation time of Large-Scale network data.(3)This paper combines the above two works to develop a Large-Scale network comprehensive situational visual analysis system.This system proposes new multilayer layout algorithms and other methods to solve the key problems encountered in the visual analysis of Large-Scale multi-layer networks.The system can quickly present the layout results of Large-Scale multi-layer networks,and realize the visual analysis of the overall situation of multi-layer networks;the analysis of the changing behavior of attributes,communication,and links of network target nodes;the analysis and screening of key network nodes and Visual analysis and other functions combined with situational awareness algorithms.Using this system,researchers can smoothly and interactively analyze Large-Scale multi-layer network data.