摘要
[研究目的]面向在线社交网络舆情数据,基于符号网络建模舆情实体的正负关系,研究符号网络社团检测模型挖掘舆情数据中的语义社团,进一步推演出对于某事件群众的普遍看法,适用于真实社会网络中的事件检测,有助于创新舆情情报分析方法。[研究方法]主要从符号网络特性入手,在半非负分解过程中引入深度学习的框架,提出深度半非负矩阵分解模型(DSNMF),进一步将舆情情报分析及复杂网络科学有机结合,利用“‘司马3忌’举报韩红爱心慈善基金会”热点事件所产生的微博舆情数据构建舆情情感符号网络,基于DSNMF模型进行舆情情报实证分析。[研究结论]大量实验表明:DSNMF模型有效提升了符号网络社团检测性能;证实了符号网络社团检测模型在舆情情报分析中的有效性。
[Research purpose]For public opinion data of online social networks,the positive and negative relationships about public subjects are modelled with signed networks.Semantic communities in public opinion data are mined by signed networks community detection model.And then the general opinions of the masses on an event can be further deduced.It is suitable for event detection in real social networks,and can become one of the important means of public opinion intelligence analysis.[Research method]Based on the characteristics of signed network,the framework of deep learning is introduced into the semi-non-negative matrix factorization model,and the deep semi-non-negative matrix factorization model(DSNMF)is proposed.This paper further combines the analysis of public opinion and complex network science,uses the Weibo public opinion data generated by a hot issue of"Han Hong Love Charity Fundation Reported by'Sima 3Ji'for Fraudulent Practice"to construct public opinion emotional signed network,and applied the proposed DSNMF to the public opinion emotional signed networks for an empirical research.[Research conclusion]Experimens show that:the proposed DSNMF improves the performance of singed networks community detection effectively;the empirical experiment results verifies the effectiveness of the proposed DSNMF in the analysis of public opinion intelligence.
作者
余韦
章金楠
朱梦丽
王佳桐
穆荣健
Yu Wei;Zhang Jinnan;Zhu Mengli;Wang Jiatong;Mu Rongjian(College of International Business,Zhejiang Yuexiu University,Shaoxing 312069;Shaoxing Key Laboratory for Smart Society Monitoring,Prevention&Control,Shaoxing 312069;College of Intelligence and Computing,Tianjin University,Tianjin 300350)
出处
《情报杂志》
CSSCI
北大核心
2022年第5期55-60,161,共7页
Journal of Intelligence
基金
国家自然科学基金青年项目“时序网络跨尺度耦合演化建模研究”(编号:62102262)研究成果之一
2021年浙江省大学生科技创新活动计划暨新苗人才计划项目“浙江省乡村数字化风险智能预警理论体系研究”(编号:2021R433009)研究成果之一。
关键词
舆情情报分析
符号网络
深度学习
社团检测
社交网络
非负矩阵分解
微博
public opinion intelligence analysis
signed network
deep learning
community detection
social network
non-negative matrix factorization
Weibo
