随着互联网的普及和不断发展，用户通过多个社交网络进行社交活动，使用社交网络带来的丰富内容和服务.通过识别出不同社交网络上的同一用户，可以有助于进行用户推荐、行为分析、影响力最大化.已有方法主要基于用户的结构特征和属性特征来识别匹配用户，大多仅考虑局部结构，且受已知匹配用户数量的限制，提出一种基于全视角特征结合众包的跨社交网络用户识别方法（overall and crowdsourced user identification algorithm，简称OCSA）.首先，利用众包提高已知匹配用户的数量；然后，应用全视角特征评价用户的相似度，以提升用户匹配的准确性；最后，利用两阶段的迭代式匹配方法完成用户识别工作.实验结果表明：该算法可显著提高用户识别的召回率和准确率，并解决了已知匹配用户数量不足时的识别问题.

With the popularity and development of Internet, people like to take part in multiple social networks to enjoy different kinds of services. Consequently, an important task is to identify users in the networks, which is helpful for user recommendation, behavior analysis and impact maximization. Most state-of-the-art works on this issue are mainly based on the user's structure features and attribute features. They prefer to exploit user's local features and are limited by the number of the known matching users. In this paper, a method based on global view features is proposed to align users with crowdsourcing (OCSA). First, crowdsourcing is used to increase the number of known matching users on networks. Then, global view features are used to evaluate the similarity between users to improve the accuracy of user identification. Finally, an iterative two-stage matching method is put forward to answer the user identification. The results of experiments show that the presented method has better performance on precision and recall, especially when the number of known matching users is insufficient.

TP311

国家自然科学基金（61472070，61672142）；国家重点基础研究发展计划（973）（2012CB316201）