摘要
微博虚假消息的判别是微博研究中的难点问题。为了实现快速准确识别,从源微博的评论角度出发定义了三个不同特征:支持性、置信度、内容相关性。利用所选三个特征作为输入,构建SVM分类算法判别消息真伪。以抓取的新浪微博上的真实数据集作为实验对象,利用提出的模型进行了实验并与人工神经网络对比,在虚假微博的识别中初步取得了较好的结果,可以有效的识别虚假消息。
Identification of the rumor information has become a challenge topice in the field of "Weibo" research. This paper defined three features of Weibo users' reviews which include support value, confidence and correlation in order to identify water army quickly and exactly. This research used these three features as the input of SVM classification algorithm to identify the truthfulness of one tweet. The experiment dataset included real user data which was abstracted from Sina Weibo. After experiments using model this paper refers to, and comparing our method with Neu- ral Network, better results are preliminarily got in the recognition of fake Weibo contents, which is very effective.
出处
《计算机仿真》
CSCD
北大核心
2016年第1期386-390,412,共6页
Computer Simulation
基金
教育部人文社会科学研究青年基金项目(13YJC860006)