Research And Implementation Of Fast Visualization Methods For Massive Graph Data

The visualization of massive graph data is one of the difficulties and hotspots of domestic and international research,and the rapid development of big data in recent years,for example,the data scale of search engine knowledge graph and large e-commerce knowledge graph has exceeded the terabyte level,so it is urgent to solve the visualization of such massive level data.The visualization of massive graph data can be divided into two sub-problems from the perspective of data presentation and data analysis: one is for the global presentation,the problem of slow rendering of the first screen of massive graph data;the other is for the local node family,which makes data analysis difficult due to the visual disorder of massive graph data.In the engineering development practice,we have encountered the following practical problems: first,when the amount of displayed node data reaches 10,000,the existing popular visualization methods cannot render quickly,and the first screen lags for more than 3 seconds;second,the data is seriously disordered after rendering,and further data analysis cannot be performed;after thorough investigation of the existing mainstream Echarts,D3,G6,etc.,it is found that the above problems cannot be solved simultaneously.The above problems cannot be solved at the same time.Based on the above facts,this paper carries out the research on the fast visualization method of massive graph data,proposes the Edge Bundling Optimization(EBO)algorithm for solving the visualization disorder and the fast visualization scheme of massive graph data based on WebGL technology,and finally breaks through the main technical difficulties of this project.In this thesis,two solutions are proposed around the above two sub-problems,namely the layout algorithm for massive graph data analysis and the fast layout method based on WebGL,and a visualization prototype system is designed based on the two methods to verify the feasibility of the method.The specific research contents are as follows:1.To address the problem of visual confusion in the visualization of massive graph data,this thesis proposes an EBO algorithm for massive graph data analysis.The algorithm improves the traditional edge bundling algorithm by constructing kd trees through spatial partitioning to speed up the retrieval of neighboring edges;meanwhile,it uses the labeling method and constructs evaluation metrics to reduce the number of iterative computations of edges and shorten the computation time.In addition,curve optimization is performed in curve plotting,and B-sample curves are used to further improve the plotting effect.Through testing with multiple data sets,the EBO algorithm shows better performance than the traditional edge bundling algorithm and other layout algorithms in several aspects,including bundling effect and time consumption.2.This thesis proposes a solution for fast rendering of massive graph data based on WebGL technology,which solves the problem of slow rendering of the first screen.This solution is divided into four steps,which focuses on using the improved hybrid layout algorithm for fast layout.The improved hybrid layout combines the advantages of the stable compatibility of force-guided layout and the speed of circular layout;at the same time,it incorporates the idea of asymptotic,incremental rendering when the data volume exceeds 10,000,which further reduces the rendering time of the first screen under massive data.In addition,this method optimizes the rendering effect by using advanced WebGL technology to realize a fog-like visual effect that can be expanded horizontally and vertically.Through a series of experiments,the first-screen rendering speed of the massive graph data visualization scheme based on WebGL technology is significantly better than other visualization schemes,while ensuring high-quality visualization effects.3.This thesis designs and implements a prototype system for visualizing massive graph data to verify the feasibility of the studied scheme.The system consists of three modules: data import,data presentation,and data analysis.After detailed system design,it solves the problem of visualizing massive graph data from two perspectives: data presentation and data analysis,considering ease of use,reliability,maintainability,and scalability.After testing,the system can complete the rapid presentation of massive graph data(10,000 nodes and 100,000 edges)within 2 seconds and interact with it within 3 seconds;meanwhile,the EBO algorithm successfully solves the visualization disorder of massive graph data and effectively improves the visual interaction and analysis capability.