Abstract: Graph visualization as an effective technology to understand the graph structure and reveal hidden self-organization is of great significance. Meanwhile, detecting hierarchical community structures in contact graph data may give reorganizational insight of complex network relationships. This paper introduces a Neo4j-based implementation of Louvain method to produce multi-level clusters, and a prototype system for graph visualization. In the system, the hierarchical structure data are mapped to a 3D botanical tree, and provide the flexible, intuitive operation to explore the potential information. Visual analysis of experimental results show that the proposed method not only exhibits sophisticated hierarchical community structures clearly, but also displays the data details variously. As a result, the method which is applied virtual reality technique provides strong technical support for graph mining.

Key words: graph data, hierarchical community structure, 3D visualization, Neo4j

摘要： 提出一种图数据的三维树形可视化方法，基于Louvain算法对图数据中的复杂的网络关系进行层次聚类，利用三维树形映射表达聚类结果，直观展示隐含于图数据中的结构关系，通过在三维场景中旋转、缩放、移动、拾取高亮等交互操作多视角地展示数据。集成开源图数据库Neo4j研发原型系统，并开展案例数据实验。实验结果表明，该方法不仅能够简洁灵活地展示图数据的总体层次结构，还能够多样化地表达数据细节，为利用虚拟现实技术探索图数据的潜在信息提供有效的技术支持。

关键词: 图数据, 层次社区结构, 三维可视化, Neo4j