User Identity Linkage Across Multiple Social Network Based on Profile Features

The same user registers on different social platforms which scatters user data across multiple platforms, these data are incomplete, unreliable, and underutilized. Through the analysis of these cross-platform data, identify different accounts corresponding to the same user's real identity. Linking cross-platform user identities can help build detailed user profiles, recommendation systems and link predictions etc. Our research divides user basic attributes into three types: string attributes, short text attributes and semi-structured attributes, according to the characteristics of different attributes, we study the calculation method of similarity between attributes. In order to improve the effect of user identity recognition, this study selects the optimal machine learning algorithm for user characteristics of different dimensions and completes the learning process by constructing and integrating multiple models. Experimental results show that the integration model based on user profile features has a certain degree of user identity link ability, can effectively integrate multiple categories of attributes between platforms, and improve the accuracy of identity link algorithm. This paper provides a reference for the research on user identification across multiple social networks.