Link Prediction in Evolving Networks Based on Popularity of Nodes

Area of Science:

• Network Science
• Data Mining
• Computational Social Science

Background:

• Link prediction is crucial for understanding network dynamics and identifying missing or spurious connections.
• Traditional methods often fail in evolving networks due to their inability to capture time-varying features.
• Node popularity, or activeness, is hypothesized to significantly influence future link formation.

Purpose of the Study:

• To develop a novel link prediction approach that incorporates both structural importance and current node popularity.
• To address the limitations of static-structure based methods in dynamic network environments.
• To enhance the accuracy and robustness of link prediction in evolving networks.

Main Methods:

• Proposed the popularity based structural perturbation method (PBSPM) to model link likelihood.
• Developed a fast algorithm for the PBSPM to efficiently process large-scale networks.
• Evaluated the method on six diverse evolving networks.

Main Results:

• The PBSPM significantly outperformed state-of-the-art methods in accuracy and robustness.
• Experimental results demonstrated that active nodes are more likely to form future links.
• Visual and statistical analyses confirmed the method's effectiveness in capturing temporal dynamics.

Conclusions:

• Node popularity is a critical factor in link prediction for evolving networks.
• The proposed PBSPM offers a more accurate and robust approach to link prediction.
• This method has implications for understanding and predicting behavior in dynamic network systems.