Investigation Into Information Search And Visual Complexity Of Edge Bundling

Network associations are the most common relationships in big data era,such as the Internet,social networks,and transportation networks.Hierarchical data structure is a special case of network data.Network visualization aims at intuitively displaying underlying relationship patterns in the network,such as node or edge aggregation,based on the topological relations of network nodes and connections.How to visualize in large-scale networks with massive amount of nodes and edges on a limited screen space is challenging.In addition to the visualization of static network topology,large data-related networks tend to feature dynamic evolution,therefore,how to visualize the characteristics of dynamic networks is also an indispensable research issue.Visual clutters have seriously hindered the understandability of graphs,resulting in a large number of aggregations and overlapping phenomenon.The straight-line diagram reduces the aesthetic quality and efficiency of information search,making it difficult for the user to distinguish the visual effects.Edge bundling provides a way to solve this problem.The purpose of edge bundling is to mine information and gain insight into the contained knowledge therein from the results of data visualization.It aggregates similar edges and forms the so-called bundles according to the level of detail principle of information visualization.These bundles,which can be recognized independently,reduces visual complexity and reveals high-level structure,i.e.skeleton or backbones,which is crucial for understanding the underlying pattern of data.Several edge bundling approaches have been published in the literature,based on diverse technique paradigms,e.g.geometry structure,force-directed model,hierarchical structure,image skeleton extraction,etc.The results from existing approaches generally show strong edge attraction,bundles with unnecessary curvature and tangling at bundle intersections.No previous experimental evaluation approach has been proposed for edge bundling.We designed a user contrast experiment to evaluate the influence of edge bundling on cognitive behaviors in the process of user data analysis in order to evaluate and compare the advantages and disadvantages of edge bundling.The whole process includes the hypothesis,subjects answered,data analysis and comparison,reporting results.The experiment includes the expert version and the trial version.First,we organized 7 graduate students who research into information visualization and are called “experts” here.They reported true or false to questions according to four versions of stimuli,against the pre-calculated ground truth.Second,two hundred graduate and undergraduate students participated in the study under the same experimental environment and hardware equipment.We empirically examine the influence of edge bundling on readability,including accuracy,efficiency and processing fluency of information search under different levels of bundling stre ngth.The paper reports our detailed findings.