Research On Visual Analytic Methods For Relational Data Based On Graph Model

The association relations exist widely in the national economy and many fields of daily life,the exploration of which is one of the great challenges for data analysis in the big data era.Graph analysis is an effective approach to reveal the complex relationships in data set,and usually visualization plays a significant role during this process.The graph-based relational data visual analytics,integrating the advantage of the analysis based on graph theory with the visualization technology,and at the same time combining together the machine intelligence and human intelligence,has become an important and effective means for analyzing various data associations.How to explore and analyze the relationship in the data intuitively and efficiently,better find and grasp the focus of the subject and track the data change,is an urgent problem to be solved.On the other hand,the level of pesticide residues is one of the significant considerations in food safety field and closely related to the human edible security and the trade among countries.Therefore,our country regularly detects pesticide residues in agricultural products every year.The associations existing in the data set are of great research value.Based on this dataset,taking“The Twelfth Five-Year” National Science and Technology Program as the background,this paper studies the data association visual analysis method based on graph,and contributions of this paper are as follows:(1)Firstly,from the perspective of visual analytical method research,according to the different characteristics of the graph layouts,the methods of data association visualization based on graph are summarized in category.The typical classification of the methods,the advantages and disadvantages of various approaches and the basic framework for the graph analysis are given in this paper.Besides,the common graph analysis methods,the evaluation standards,the interaction techniques and graph simplification technologies used in the graph analysis process are also talked about.(2)A transformation-based visualization method called TransGraph is proposed.Given the problem of visual clutter,high complexity of layout and that the key objects are not given enough attention during the analysis for complex associations using the traditional graph layout method,TransGraph introduces a cascade weighted bipartite graph model based on the center point,reduces the visual complexity with the transformation on the traditional graph layout approaches including force directed layout,sunburst and radial layout algorithms,and in the same time lays much more emphasis on the current key focus.(3)A novel visualization approach to explore the time-varying hierarchical data,which is called OSMT(Ordered Small Multiple Treemaps),and a new metric called TVA(Ability of Tracking time-varying data in treemap)are proposed.OSMT visualizes the time-varying hierarchical relational data by using the treemaps in a Small Multiple display,calculating the global optimal average aspect ratio for nodes on all Small Multiple snapshots,based on which the location of the nodes are controlled.Interaction techniques like zooming and highlighting are used to further improve screen utilization and reduce node location time.Besides,in order to evaluate treemaps from the aspect of difficulty level for tracking time-varying nodes,the metric TVA considering both the change of node distance and angle is proposed.(4)Finally,an interactive visual analytic system is provided for analyzing the detection and hierarchical relationships in the simulated pesticide residue data sets,the time-varying characteristics of data nodes and the potential similarity among them,where the approaches proposed above are applied practically,not only assisting the experts in the food safety field to make corresponding decisions as an effective visual analytic tool,but also providing reference for the analyzers in some other fields like literature co-citation or film cooperation analysis,etc.