摘要
提出基于深度学习BCCM模型的网上用户画像识别方法,改善以往网上用户画像识别方法仅提取行为特征导致识别精度低的缺陷。采用爬虫软件挖掘网上用户的访问量、评论量、转发量、影响力、关注量以及网龄六种行为特征,采集网上用户发帖以及评论文本信息建立用户向量,规约处理所建立用户向量。以类别的总距离平方和最小为聚类目标,采用K-means聚类算法聚类处理用户向量。设置聚类结果为网上用户内容特征,将所获取的内容特征与所挖掘行为特征输入深度卷积神经网络中,通过卷积操作以及池化操作实现网上用户画像的有效识别。实验结果表明,采用该方法识别网上用户画像的精度高于99%,F1值高于0.92,说明所提出方法的识别精度高。
Research on the recognition of online user portraits based on the deep learning BCCM model,to improve the previous recognition of online user portraits only extracting behavioral features,resulting in low recognition accuracy.Use crawler software to mine the six behavioral characteristics of online users'visits,comments,reposts,influence,attention,and Internet age,collect online user postings and comment text information to establish user vectors,and process the established user vectors by protocol processing,using K-Means clustering algorithm takes the minimum sum of squared distances of the different categories as the clustering target.After the clustering analysis protocol is processed,the user vector is set,the clustering result is set as the online user content feature,and the obtained content feature is compared with the excavated behavior feature In the input deep convolutional neural network,effective recognition of online user portraits is achieved through convolution and pooling operations.Experimental results show that the accuracy of using this method to identify online user portraits is higher than 99%,the F1 value is higher than 0.92,and the recognition accuracy is high.It can be used in applications such as network intelligent recommendation systems.
作者
王瑛
WANG Ying(Minjiang College,Fuzhou 350108,China)
出处
《齐齐哈尔大学学报（自然科学版）》
2022年第5期11-16,共6页
Journal of Qiqihar University(Natural Science Edition)
