随着微博平台的流行微博平台上产生了大量难于验证的谣言信息并在用户之间快速传播对平台的发展和社会稳定带来巨大危害。如何自动检测并识别谣言对于制止谣言传播、减小谣言危害具有极其重要的意义。基于以上背景开展了谣言检测的相关研究。本文对来源于新浪微博的谣言数据集的4664条标注数据进行研究利用谣言数据中的文本信息以及时间信息进行建模构建出门控循环单元网络(GRU)模型实现了基于序列模型的微博谣言检测算法在测试集上达到了89.36%的准确率。在此基础上利用卷积神经网络(CNN)来提取文本的深层特征构建出CNN-GRU复合网络模型并获得了测试集上92.43%的准确率。基于上述实验结果本文进一步对两种谣言检测模型进行早期谣言检测实验取得了不错的效果。关键词 微博谣言 早期检测 机器学习 深度神经网络With the popularity of the Weibo platform, a large number of rumor messages that are difficult to verify are generated on the Weibo, and spread rapidly among users, which brings great harm to the development of the Weibo platform and social stability. How to automatically detect and identify rumors is of great significance for stopping the spread of rumors and reducing the harm of rumors. Based on the above background, relevant research on rumor detection was carried out.This paper studies the 4664 annotation data from the rumor dataset of Sina Weibo, and uses the text information and time information in the rumor data to model and construct a Gated Recurrent Unit(GRU) model to detect rumors. This rumor detection algorithm achieved 89.36% accuracy on the test set. On this basis, the Convolutional Neural Network(CNN) is used to extract the deep features of the text, and the CNN-GRU composite network model is constructed and the accuracy of 92.43% on the test set is obtained. Based on the above experimental results, this paper further conducts early rumor detection experiments on two rumor detection models, and has achieved great effects.Keywords Rumors Early Detection Machine Learning Deep Neural Network目 次1 绪论…………………………………………………………………………………….………11.1 研究背景及意义……………………………………………………………….…… 11.2 相关技术的现状……………………………………………………………………… 21.3 总体技术方案及其社会影响………………………………………………………… 41.4 论文章节安排………………………………………………………………………… 42 问题描述及方法基础………………………………………………………………………… 62.1 问题描述……………………………………………………………………………… 62.2 卷积神经网络……………………………………………………………………… 62.3 门控循环单元………………………………………………………………………… 73 微博谣言检测模型…………………………………………………………………………… 103.1 数据预处理……………………………………………………………………………103.2 基于门控循环单元的网络模型……………………………………………………… 123.3 基于卷积-门控循环单元的网络模型…………………………………………………143.4 目标函数及优化方法…………………………………………………………………154 模型实验与评估………………………………………………………………………………164.1 数据集…………………………………………………………………………………164.2 早期谣言检测…………………………………………………………………………184.3 评估方法及实验参数设置……………………………………………………………184.4 实验结果分析与改进………………………………………………….………………19结论……………………………………….……………………………………………………… 24致谢.………………………………………………………….……………………………….….25参考文献…………………………………………………………………………………………26包含设计文档任务书选题表中期表外文翻译及原文参考文献周记评阅表答辩记录表Python项目源码
随着微博平台的流行微博平台上产生了大量难于验证的谣言信息并在用户之间快速传播对平台的发展和社会稳定带来巨大危害。如何自动检测并识别谣言对于制止谣言传播、减小谣言危害具有极其重要的意义。基于以上背景开展了谣言检测的相关研究。本文对来源于新浪微博的谣言数据集的4664条标注数据进行研究利用谣言数据中的文本信息以及时间信息进行建模构建出门控循环单元网络(GRU)模型实现了基于序列模型的微博谣言检测算法在测试集上达到了89.36%的准确率。在此基础上利用卷积神经网络(CNN)来提取文本的深层特征构建出CNN-GRU复合网络模型并获得了测试集上92.43%的准确率。基于上述实验结果本文进一步对两种谣言检测模型进行早期谣言检测实验取得了不错的效果。关键词 微博谣言 早期检测 机器学习 深度神经网络With the popularity of the Weibo platform, a large number of rumor messages that are difficult to verify are generated on the Weibo, and spread rapidly among users, which brings great harm to the development of the Weibo platform and social stability. How to automatically detect and identify rumors is of great significance for stopping the spread of rumors and reducing the harm of rumors. Based on the above background, relevant research on rumor detection was carried out.This paper studies the 4664 annotation data from the rumor dataset of Sina Weibo, and uses the text information and time information in the rumor data to model and construct a Gated Recurrent Unit(GRU) model to detect rumors. This rumor detection algorithm achieved 89.36% accuracy on the test set. On this basis, the Convolutional Neural Network(CNN) is used to extract the deep features of the text, and the CNN-GRU composite network model is constructed and the accuracy of 92.43% on the test set is obtained. Based on the above experimental results, this paper further conducts early rumor detection experiments on two rumor detection models, and has achieved great effects.Keywords Rumors Early Detection Machine Learning Deep Neural Network目 次1 绪论…………………………………………………………………………………….………11.1 研究背景及意义……………………………………………………………….…… 11.2 相关技术的现状……………………………………………………………………… 21.3 总体技术方案及其社会影响………………………………………………………… 41.4 论文章节安排………………………………………………………………………… 42 问题描述及方法基础………………………………………………………………………… 62.1 问题描述……………………………………………………………………………… 62.2 卷积神经网络……………………………………………………………………… 62.3 门控循环单元………………………………………………………………………… 73 微博谣言检测模型…………………………………………………………………………… 103.1 数据预处理……………………………………………………………………………103.2 基于门控循环单元的网络模型……………………………………………………… 123.3 基于卷积-门控循环单元的网络模型…………………………………………………143.4 目标函数及优化方法…………………………………………………………………154 模型实验与评估………………………………………………………………………………164.1 数据集…………………………………………………………………………………164.2 早期谣言检测…………………………………………………………………………184.3 评估方法及实验参数设置……………………………………………………………184.4 实验结果分析与改进………………………………………………….………………19结论……………………………………….……………………………………………………… 24致谢.………………………………………………………….……………………………….….25参考文献…………………………………………………………………………………………26包含设计文档任务书选题表中期表外文翻译及原文参考文献周记评阅表答辩记录表Python项目源码
