Temporal walk based centrality metric for graph streams.

Area of Science:

• Network Science
• Data Mining
• Computational Social Science

Background:

• Evaluating node importance in networks is challenging, especially in dynamic networks like Twitter.
• Existing centrality measures struggle with the fast-paced nature of real-world network evolution.
• Predicting emerging centrality in dynamic networks remains a significant research problem.

Purpose of the Study:

• To propose a novel temporal walk-based dynamic centrality measure.
• To model temporal information propagation by incorporating the order of edge creation.
• To quantitatively assess and compare the performance of dynamic centrality metrics.

Main Methods:

• Developed a new dynamic centrality measure based on temporal walks.
• Designed a quantitative experiment to evaluate temporal centrality metrics.
• Compared the proposed measure against static and other dynamic centrality approaches.

Main Results:

• The proposed temporal walk-based measure outperforms static and other dynamic measures.
• It effectively assigns higher time-aware centrality to relevant nodes in dynamic networks.
• The method demonstrates proficiency in detecting concept drift in graph generation processes.

Conclusions:

• The new temporal centrality measure offers a more accurate way to assess node importance in dynamic networks.
• This approach enhances our understanding of information propagation and network evolution.
• The method has practical applications in identifying key influencers and understanding network changes over time.