基于EM-LDA综合模型的电商微博热点话题发现

发布时间：2018-04-18 20:41

本文选题：电商微博 + 热点话题挖掘　；参考：《现代图书情报技术》2015年11期

【摘要】：【目的】在社交营销环境下,准确且有效地挖掘电商微博中的热点话题。【方法】提出一种综合模型EM-LDA对电商微博文本数据进行主题挖掘。EM-LDA综合模型包含两个子模型:ET-LDA模型和IT-LDA模型,前者对含有哈希标签的微博进行主题挖掘,后者对不含有哈希标签的微博进行主题挖掘。【结果】在确定合适的主题个数之后,标准LDA模型和EM-LDA综合模型均被用来挖掘电商微博文本数据的热点话题,与标准LDA模型相比,EM-LDA综合模型的热词挖掘准确率和有效性均较高,且能提高主题可解释性。【局限】在ET-LDA模型中,未考虑微博联系人之间的关联关系,即模型中未引入用户特征;在IT-LDA模型中没有考虑如何处理那些既是转发式又是对话式的电商微博。【结论】EM-LDA综合模型根据数据的特点,改进了标准LDA模型,能够提升电商微博热点话题识别的准确性。
[Abstract]:[objective] in the context of social marketing,Accurate and effective mining of hot topics in e-commerce Weibo. [methods] A synthesis model named EM-LDA is proposed for topic mining. EM-LDA synthesis model consists of two sub-models:: ET-LDA model and IT-LDA model.The former carries on topic mining for Weibo with hash tag, while the latter carries out topic mining for Weibo without hash tag. [results] after determining the appropriate number of topics,Both the standard LDA model and the EM-LDA synthesis model are used to mine hot topics of e-commerce Weibo text data. Compared with the standard LDA model, the EM-LDA synthesis model has higher accuracy and effectiveness of hot word mining.And can improve the theme interpretability. [limitations] in the ET-LDA model, did not take into account the relationship between Weibo contacts, that is, the model does not introduce user characteristics;The IT-LDA model does not consider how to deal with the electronic commerce Weibo, which is both forwarding and conversational. [conclusion] according to the characteristics of the data, the EM-LDA synthesis model improves the standard LDA model, which can improve the accuracy of the identification of the hot topics in the ecommerce Weibo model.
【作者单位】：东南大学经济管理学院;江苏经贸职业技术学院信息技术学院;
【基金】：江苏省高校哲学与社会科学重点项目“江苏网络经济发展现状与对策研究”(项目编号:2013ZDIXM017)的研究成果之一
【分类号】：TP391.1

【参考文献】