基于聚类分析的网络舆情主题提取

Extracting topic of network public opinion with clustering analysis

网络舆情具有时效性强、传播迅速、涉及方面杂而广、意见指向性特征明显、泛娱化特征明显等特点。因此,提出对LDA输入数据采用TF-IDF算法加强特征词筛选的方法。选取"巴黎圣母院大火"事件,采集作为网络舆情重要来源的微博数据,进行LDA建模,引入TF-IDF算法进行特征词的筛选,能较准确地分析出该事件的主题分布。

Network public opinion has the characteristics of strong timeliness, rapid dissemination, wide and miscellaneous involved aspect, obvious directional feature of opinion, and obvious characteristics of pan-entertainment. Therefore, TF-IDF algorithm is proposed to enhance the feature word filtering for LDA input data. Select "Notre Dame Fire in Paris" event, collect micro-blog data, which is an important source of public opinion on the network, model it with LDA, and introduce TF-IDF algorithm to select the feature words, it is found that the subject distribution of the event can be analyzed more accurately.