基于深度学习的垃圾邮件检测  

Spam Detection Based on Deep Learning

作　　者：俞荧妹 禹素萍[1] 许武军[1] 范红[1] 

机构地区：[1]东华大学信息科学与技术学院,上海

出　　处：《计算机科学与应用》2023年第4期764-772,共9页Computer Science and Application

摘　　要：邮件是日常生活中的一种通讯工具,但垃圾邮件对用户造成严重困扰,因此改进垃圾邮件识别技术、提升其准确率和效率具有重要现实意义。在文本分类领域,深度学习有很好的应用效果。故文章提出了一种基于CNN的BiGRU-Attention模型,旨在充分利用CNN的特征提取能力和BiGRU的全局特征提取能力。引入注意力机制能够突出显示重要文本,前后共经过两层双向门控循环单元,从而更全面地提取邮件文本特征。实验数据选取Trec06c数据集,并与其他分类模型对比,结果表明,检测准确率达到91.56%。Email is a communication tool in daily life, but spam has caused serious problems for users, As a result, it is crucial to improve spam identification technology and improve its accuracy and efficiency. In the field of text classification, deep learning has a good application effect. In order to fully utilize CNN’s feature extraction capabilities and BiGRU’s global feature extraction capabilities, this article suggests a CNN-based BiGRU-Attention model. The introduction of the attention mechanism can highlight important text, which passes through two layers of two-way gated loop units before and after, so as to extract more comprehensive features of email text. The experimental data is selected from Trec06c dataset and compared with other classification models. The results show that the detection accuracy reaches 91.56%.

关 键 词：垃圾邮件 文本分类 深度学习 双向门控循环单元 注意力机制 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]