摘要
近年来,许多社交网络平台为保护用户的隐私限制了网络结构获取API的访问频率,这给跨社交网络的用户识别带来了新的挑战.针对此问题,本文提出一种融合多特征的跨社交网络用户在线识别方法(Multi-feature Cross-social network User Online Recognition method,MCUOR).该方法改进了属性相似度计算方法,设计了动态爬虫策略DYN实现在API限制范围内获取更多有效结构信息,结合好友权重改进了加权Jaccard相似度从而改进了局部结构特征的提取方法,最后利用逻辑回归模型结合属性特征和局部结构特征构建用户识别模型对用户进行跨社交网络用户在线识别.真实数据集上的实验结果表明爬虫策略DYN实现了有限API的优质调配,与其他方法相比MCUOR方法提高了用户识别精确度和召回率.
In recent years,many social networking platforms limit the access frequency of network structure acquisition API to protect users'privacy,which brings new challenges to cross-social network user recognition.To solve this problem,we propose multi-feature online recognition method which improves the attribute similarity calculation method,designs a dynamic crawler strategy DYN to obtain more effective structure information within the limited API,and improves the weighted Jaccard similarity coefficient by combining the friend weight to improve the local structure feature extraction method.Finally,a logistic regression model is used to combine attribute features and structural features to construct a user recognition model to identify users across social networks.The experimental results on the real data set show that the crawler strategy DYN achieves high-quality deployment of limited APIs.Compared with other methods,the MCUOR method improves the precision and recall of user recognition.
作者
卢菁
王菊钿
刘丛
LU Jing;WANG Ju-tian;LIU Cong(School of Optical-electrical and Computer Engineering,University of Shanghai for Science&Technology,Shanghai 200093,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2021年第11期2407-2414,共8页
Journal of Chinese Computer Systems
基金
国家自然科学基金青年基金项目(61703278)资助
上海理工大学自然基金培育项目(20ZRPY08)资助.
关键词
跨社交网络
用户识别
爬虫策略
逻辑回归模型
特征提取
cross-social network
user recognition
crawler strategy
logistic regression model
feature extraction
