基于标题聚类的论坛舆论领袖发现算法  

Opinion leader discover based on title clustering in forum

作　　者：周雪妍[1,2] 杨静[2] 林泽鸿[1] 吉亚力[3] 

机构地区：[1]哈尔滨学院工学院,黑龙江哈尔滨150086 [2]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150000 [3]哈尔滨师范大学管理学院,黑龙江哈尔滨150025

出　　处：《计算机工程与设计》2014年第12期4316-4319,4334,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61370083;61073043);高等学校博士学科点专项科研基金项目(20122304110012);哈尔滨市青年后备人才专项基金项目(2014RFQXJ081);黑龙江省教育科学研究青年专项课题基金项目(GBD1213045);教育厅人文社科基金项目(12542083);哈尔滨师范大学人文社科预研基金项目(SYB2012-02);哈尔滨学院学科发展青年基金项目(HUYF2013-011)

摘　　要：为进行论坛舆情分析,提出一种基于标题聚类的舆论领袖发现算法。按时间将数据进行预处理,运用话题模型度量标题数据并依此进行标题聚类;建立同一话题下的变规模用户回复关系网络,结合情感分析和网络特性分析进行影响力排名以提取舆论领袖。该算法旨在快速发现某一网络热门事件中的舆论领袖,综合考虑了帖子的话题属性、情感倾向和网络结构关系。通过实验验证了该模型的可行性和有效性。Aiming at the problems of the forum public opinion analysis,an opinion leader discover algorithm was provided,which effectively utilized the title clustering to identify the opinion leader in the fields of interest.The post titles were preprocessed firstly based on their publication date and then clustered based on the semantic model.In the end,variable scale posts reply relational networks were set up for the social network analysis and the sentiment analysis,and the users were ranked to identify the opinion leaders.Algorithm is designed to find the opinion leader in a hot event of the network quickly considering topic attributes,sentiment orientation and network structure.The feasibility and effectiveness of the model is verified by experiments.

关 键 词：论坛 舆论领袖 标题聚类 情感分析 舆情分析 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]