一种加权图卷积神经网络的新浪微博谣言检测方法  

Sina Microblog Rumor Detection Method Based on Weighted-graph Convolutional Network

作　　者：王昕岩 宋玉蓉 宋波[2] WANG Xin-yan;SONG Yu-rong;SONG Bo(College of Automation&College of Artificial Intelligence,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;School of Modern Posts&Institute of Modern Posts,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学自动化学院、人工智能学院,南京210023 [2]南京邮电大学现代邮政学院,南京210003

出　　处：《小型微型计算机系统》2021年第8期1780-1786,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61672298)资助;江苏高校哲学社会科学研究重点项目(2018SJZDI142)资助;教育部人文社会科学研究规划基金项目(17YJAZH071)资助。

摘　　要：信息化时代下,广泛传播的谣言极大地影响了人们的日常生活,甚至威胁了社会稳定,因此对谣言的检测任务具有现实意义.目前基于深度学习模型的谣言检测方法忽略了事件之间的联系或事件之间联系的紧密程度,对检测效果造成了一定影响.本文考虑事件之间联系的异质性,将事件之间联系的紧密程度描述为连边权重,提出了一种基于加权图卷积神经网络(Weighted-Graph Convolutional Netw ork,W-GCN)模型的新浪微博谣言检测方法.该方法通过W-GCN模型,学习得到节点的隐层表示,进而对节点进行分类,最终完成谣言检测任务.实验结果表明,与现有的谣言检测方法相比,本文提出的基于W-GCN模型的谣言检测方法,可以提高谣言检测的正确率、精确率、召回率和F1值,即能更有效地识别谣言.In the era of information,widespread rumors greatly affect the daily life of people and even threaten social stability.So it has practical significance for the task of rumor detection.At present,rumor detection methods based on the deep learning model ignore the connection between events or the closeness of the connection,which affects the detection results in a certain degree.Considering the heterogeneity of the connection between events,describing the closeness of the connection as the weight of the edge,we propose a new Sina Microblog Rumor Detection method based on Weighted-Graph Convolutional Network(W-GCN).In this method,we use W-GCN model to study the hidden representation of the nodes,and then classifies the nodes to finally complete the rumor detection task.Experiment results show that,compared with the existing rumor detection methods,the rumor detection method based on W-GCN model we proposed can improve the accuracy,precision,recall and F1-measure of rumor detection,that is to say,it can be more effective for rumor identification.

关 键 词：图卷积神经网络 连边权重 新浪微博 谣言检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]