摘要
互联网正在成为舆论的传播平台。重要的是要尽可能准确地模拟互联网舆论活动。对谣言、假新闻、误导的信息与不正确的信息等网络舆情的监测,是解决当今网络安全问题的关键,因为上述信息的传播可能会对我们的社会稳定发展产生严重的后果。为了解决该问题,提出一种基于LSTM(长短期记忆)的深度学习的社会网络舆情监测。该模型使用Word2Vec算法中的CBOW模型,该模型能将单词序列转换为向量序列,然后将向量序列输入到LSTM模型中。最后,在LSTM模型的最后一个时间输出的预测类作为舆情监测的判断依据。实验结果表明,本文在舆情监测上提出的模型在精准度、召回率和F1分数等方面优于其他先进的网络舆情监测方法。本文方法的准确率较本实验中表现最好的方法提升10%,且时效性大大增加。
The Internet is becoming a dissemination platform of public opinion.It is important to simulate Internet public opinion activities as accu⁃rately as possible.The monitoring of online public opinion such as rumors,fake news,misleading information,and incorrect information is the key to solving today's network security problems,because the dissemination of the above information may have serious consequences for the stable development of our society.In order to solve this problem,this paper proposes a social network public opinion monitoring based on LSTM(Long Short Term Memory)deep learning.The model uses the CBOW model in the Word2Vec algorithm,which can con⁃vert the word sequence into a vector sequence,and then input the vector sequence into the LSTM model.Finally,the prediction class out⁃put at the last time of the LSTM model is used as the judgment basis for public opinion monitoring.Experimental results show that the mod⁃el proposed in this article is superior to other advanced online public opinion monitoring methods in terms of accuracy,recall rate and F1 score.The accuracy of this method is 10%higher than the best method in this experiment,and the timeliness is greatly increased.
作者
王民昆
王浩
苏博
WANG Min-kun;WANG Hao;SU Bo(Southwest Branch of State Grid Corporation of China,Chengdu 610041;College of Software Engineering,Chengdu Information Technology University,Chengdu 610225)
出处
《现代计算机》
2020年第33期20-24,共5页
Modern Computer
