模式无关的社交网络用户识别算法

发布时间：2018-04-09 08:37

本文选题：用户识别　切入点：二部图　出处：《西安交通大学学报》2013年12期

【摘要】：针对识别社交网络用户时存在的模式不一致问题,提出了基于分块和二部图的用户识别算法。该算法通过将传统分块算法中的属性值精确匹配扩展为无模式信息下的属性值近似匹配,避免了传统用户识别时所需的模式对齐;使用加权二部图及Kuhn Munkres(KM)最大权匹配算法进行源用户档案与待匹配用户档案间的相似度计算,解决了用户档案间属性个数不同及语义语法异构的问题。在社交网站Profilactic上采集了965个用户的公开数据,采用召回率、精确率和综合指标等评价指标对算法进行了实验评估。实验结果表明,所提算法能够不依赖模式信息进行实例级跨系统用户识别,与基于属性值精确匹配的算法相比,所提算法的召回率提高了6.2%~9.5%,综合评价指标提高了3%~4.2%。
[Abstract]:In order to solve the problem of pattern inconsistency in the identification of social network users, a user identification algorithm based on block and bipartite graph is proposed.By extending the exact matching of attribute values in the traditional block algorithm to approximate matching of attribute values without schema information, the proposed algorithm avoids the pattern alignment required by traditional user identification.By using weighted bipartite graph and Kuhn Munkres KM) maximum weight matching algorithm, the similarity between the source user file and the user file to be matched is calculated, which solves the problem of different number of attributes and heterogeneous semantic syntax among user files.The public data of 965 users were collected on the social network Profilactic. The algorithm was evaluated experimentally by using the evaluation indexes such as recall rate, precision rate and comprehensive index.The experimental results show that the proposed algorithm can recognize cross-system users without depending on the pattern information. Compared with the algorithm based on the exact matching of attribute values, the recall rate of the proposed algorithm is increased by 6.2and 9.5. the comprehensive evaluation index is increased by 34.2.
【作者单位】：西安交通大学电子与信息工程学院;西安建筑科技大学信息与控制工程学院;陕西电力信通有限公司;
【基金】：国家自然科学基金资助项目(61272458) 陕西省自然科学基础研究计划资助项目(2013JM8021) 西安建筑科技大学青年基金资助项目(2013JK1189)
【分类号】：TP393.09

【相似文献】