Abstract: Hotspots detection is one of the important tasks of the literature analysis, which would help researchers to understand an unfamiliar trans-disciplinary. From the perspective of graph mining, this research makes use of the complex network community detection method to detect research hotspots in domestic forestry of scientific and technical literature. At the same time, this research analyses the associations between hotspots. In the research, the forestry literatures from 2000 to 2012 are taken as the research object, and a co-word weighted network is constructed. Then the network topology characteristics are analyzed from community structural perspective. The result shows that the network is endowed with small-world property, scale-free nature and eight communities. These eight communities correspond to the forestry field respectively, ie. the eight main research fields. This research is shown in a visual way so that each topic domain consisting of series of hotspot can be seen, so is their relationships. Hence the network analysis can be used as a core method of literature recommending system.

Key words: forestry literature, complex network, community detection, big data, graph mining, co-word network, visualization

摘要： 热点识别是文献分析的重要任务之一，针对专业领域的文献分析有助于研究人员快速掌握相关领域的核心问题。从图挖掘的角度利用复杂网络的社区识别技术来识别我国的林业科技文献的研究热点，并且分析这些热点之间的关联关系。以2000—2012年林业科技文献为研究对象，构建了一个文献关键词共现加权网络，从社区结构的角度分析了该网络的拓扑特性。结果显示该网络具有明显的小世界特征、无标度特征和八大社区结构。这八个社区分别对应着林业领域的八大主题研究领域，以可视化的方式展示了每个主题域由一系列的热点组成，主题域之间也呈现了疏密的关系。因此，所采用的网络分析方法可以作为科技文献推荐系统的核心方法。

关键词: 林业科技文献, 复杂网络, 社区识别, 大数据, 图挖掘, 关键词共现网络, 可视化