It is difficult to find the microblog hotspot because the characteristics of microblog are short,rapid,change and so on.This paper proposed a microblog hotspot detection method based on FIHC and TOPSIS in order to solve the problem.Firstly,it used the calculation of HowNet similarity into the score function of FIHC,and considered the semantic links between frequent words,and produced the initial clusters based on frequent words more accurately.Then it reduced the initial cluster of the text repletion of mircoblog,and used the idea of single-pass clustering to the reduced topic cluster in order to get the hotspot.At last,it used an improved TOPSIS model to sort the hot topics in order to get the rank of the hot topics.Comparing with the other text clustering algorithms and hotspot detection methods,the method has good effect,and can be more comprehensive response to the current hot topics.%针对微博的文本存在短小、快速、变化等特点,导致热点发现困难的问题,提出了一种基于改进的FIHC聚类和TOPSIS的热点发现方法.首先把知网语义相似度引入FIHC聚类算法score函数的计算,考虑了频繁词之间的语义联系,更准确地生成基于频繁词的初始簇;然后对微博文本重复的初始簇进行消减,再采用Single-Pass聚类的思想对消减完的话题簇进一步聚类最终得到热点话题;最后对热点话题采用改进的TOPSIS模型进行排序,更好地获得热点话题的排行.通过与其他文本聚类算法以及热点发现方法对比,该方法热点发现效果好,能够比较全面地反映当前的热点话题.
展开▼
