Friend closeness based user matching cross social networks

Area of Science:

• Computer Science
• Artificial Intelligence
• Social Network Analysis

Background:

• User matching across social networks is crucial for identifying individuals.
• Existing methods often overlook the importance of social connections and friend proximity.
• Current approaches primarily rely on user attributes and network embeddings.

Purpose of the Study:

• To develop a novel algorithm for cross-network user matching that incorporates friend closeness.
• To improve the accuracy and generalization of user matching by considering social network structures.
• To address the limitations of existing methods that ignore the influence of friends.

Main Methods:

• Proposed a friend closeness based user matching (FCUM) algorithm.
• Utilized an attention mechanism to quantify user-friend closeness.
• Developed a semi-supervised, end-to-end learning framework.
• Integrated individual similarity and close friend similarity into a single objective function.
• Implemented a bi-directional matching strategy to reduce labeling costs.

Main Results:

• FCUM demonstrated superior performance compared to state-of-the-art methods.
• The inclusion of friend closeness significantly improved matching accuracy.
• The bi-directional matching strategy proved effective in reducing training data requirements.
• Experiments on real-world datasets validated the algorithm's effectiveness.

Conclusions:

• Friend closeness is a vital factor in accurate cross-network user matching.
• The FCUM algorithm offers a robust and efficient solution for user identification across social platforms.
• The proposed method enhances generalization and reduces the need for extensive labeled data.