A Method for Analyzing User Identity Association relationships Based on Multi-level User Attributes
Yi Zhang, Shibin Zhang · 2024
In the face of the increasingly prosperous state of the Internet, it remains challenging to integrate various social network user information. The first step towards achieving data integration is to identify the account information of the same user on different platforms. This paper proposes a social network user identity correlation analysis method based on multi-level user attributes, which mainly includes user attribute embedding and regularized canonical correlation analysis. The former combines machine learning with the user's posted content and self-assigned attributes for representation embedding, while the latter maps the vector representation of the social network to a common correlation space to solve for maximum correlation, thereby determining whether two different social network users are the same user. Applying this correlation analysis method to a real-world network dataset for simulation experiments, the experimental results demonstrate that this correlation analysis method achieves higher hit accuracy compared to other traditional methods.