基于云计算的网络舆情藤蔓式串并行聚类方法研究  

Research on Vine Type Serial-Parallel Clustering Method of Network Public Opinion Based on Cloud Computing

作　　者：陈雪刚[1] Chen Xuegang(College of Software and Communication Engineering, Xiangnan University, Chenzhou 42300)

机构地区：[1]湘南学院软件与通信工程学院,郴州423000

出　　处：《情报杂志》2017年第2期122-126,107,共6页Journal of Intelligence

基　　金：教育部人文社会科学研究项目"基于大数据分析的在线社会网络舆情挖掘及风险管控研究"(编号:15YJC870004);湖南省哲学社会科学基金项目"网络环境下社会舆情危机事件的识别及评价策略研究"(编号:13YBA302)

摘　　要：[目的/意义]针对目前在线网络海量信息难以快速获取其舆情的问题,为有效提升舆情获取速度,提出一种基于云计算的网络舆情藤蔓式串并行聚类方法。[方法/过程]首先构建舆情文本相似性度量模型,并设计确定舆情最佳聚类阈值的方法,以度量舆情信息间的相似性及确定舆情最佳聚类阈值;然后构建舆情藤蔓式串并行聚类模型,以一个舆情主题为聚类起始,之后如藤蔓的主干自动分支和延伸样,自主扩展舆情聚类主题,并根据舆情信息间的相似性和聚类阈值聚类同主题的舆情信息;最后在舆情藤蔓式串并行聚类模型的基础上,利用云计算的并行计算能力,快速、准确地聚类各主题的舆情信息。[结果/结论]实验结果表明:该方法能快速获取网络舆情,且具有较高的舆情聚类准确度,可为在线网络舆情实时监控提供一定的方法支持。[ Purpose/Significance] Since it is difficult to quickly obtain the public opinion from the massive online network information, to accelerate the speed of obtaining the public opinion, the study puts forward a vine type serial-parallel clustering method of network pub- lic opinion based on cloud computing. [ Method/Process] To measure the similarity between public opinion and cluster public opinion ac- curately, the study firstly constructs the similarity measuring model of public opinion text, and designs a method which determines the opti- mal clustering threshold of public opinion, then constructs a vine type serial-parallel clustering model of public opinion, and takes a topic as the beginning of clustering, and then expands clustering theme of public opinion autonomously, and clusters the public opinion informa- tion according to its similarity and the clustering threshold, which takes the form of vine trunk branches and extends autonomously. Final- ly, on the basis of vine type serial-parallel clustering model of public opinion, using of parallel computing ability of cloud computing, net- work public opinion of each topic can be clustered quickly and accurately. [ Result/Conclusion ~ The experimental results show that the vine type serial-parallel clustering method of network public opinion based on cloud computing can quickly obtain the network public opin- ion with high accuracy, and provide a method for online real-time monitoring of network public opinion.

关 键 词：在线网络舆情 云计算 文本相似性 藤蔓式串并行聚类 

分 类 号：G202[文化科学—传播学]