Abstract:
With the increase of graph size, there appears visual clutter for the traditional force directed layout algorithms, such as node overlapping and edge crossing. Aiming to address this problem, a scalable hierarchical visual abstraction method based on improved force directed layout is proposed. We first combine the advantages of FR algorithm and LinLog algorithm to improve the force directed algorithm, which generates the preliminary layout for graphs with clear clustering structures. Based on this layout results, a bottom-up hierarchical clustering method is used to generate the graph's hierarchical structure. We define the parameter of abstraction levels to determine the clustering under different hierarchies, so that the users can observe the graph layout structures at multiple levels. Finally, we use three different metrics, which are Euclidean geometric distance, topological structure based distance and topological distance adding betweenness centrality, to compute the visual abstractions. The corresponding layout abstractions are compared and analyzed. In order to illustrate the effectiveness of the proposed method, three real-world datasets are visualized and analyzed in our experiments. The three datasets are citation data of information visualization papers, the political blog data for the 2004 US presidential campaign and the co-author data from the IEEE Visualization conference. The experimental results show that combining the visual interactive technologies such as panning, zooming and selecting, our method can help users to analyze, explore and understand the hidden information of the data effectively.