EEUPL: Towards effective and efficient user profile linkage across multiple social platforms

Linking user profiles belonging to the same people across multiple social networks underlines a wide range of applications, such as cross-platform prediction, cross-platform recommendation, and advertisement. Most of existing approaches focus on pairwise user profile linkage between two platforms, which can not effectively piece up information from three or more social platforms. Different from the previous work, we investigate scalable user profile linkage across multiple social platforms by proposing an effective and efficient model called EEUPL, which can detect duplicate profiles within one platform that belong to same person and is implemented with Apache Spark for distributed execution. The model contains two key components: 1) To link cross-platform user profiles effectively, we propose an average-link strategy based clustering method. 2) To extend the model EEUPL to large-scale datasets, an Apache Spark based approach is developed. Extensive experiments are conducted on two real-world datasets, and the results demonstrate the superiority of the model EEUPL compared with the state-of-art methods.