Using friends as sensors to detect global-scale contagious outbreaks

Area of Science:

• Computational Social Science
• Network Science
• Epidemiology

Background:

• Global-scale online data monitoring is crucial for detecting various phenomena, including epidemics and market trends.
• Privacy concerns and data scale limit traditional global monitoring; existing methods underutilize local network structures.
• Identifying key nodes for monitoring is essential for efficient early detection.

Purpose of the Study:

• To develop a model for contagious information spread in global social networks.
• To demonstrate an efficient local monitoring method for early detection of outbreaks.
• To leverage local network structure for improved monitoring.

Main Methods:

• Developed a model for contagious information spread in a global social network.
• Implemented a local monitoring strategy by selecting a small fraction of nodes and their friends.
• Analyzed six months of data from the Twittersphere to evaluate the method's effectiveness.

Main Results:

• The local monitoring group exhibited higher network centrality compared to random groups.
• Detected viral outbreaks of novel hashtags approximately 7 days earlier than random sampling.
• Central actors' increased activity and information diversity enhanced detection accuracy beyond network structure alone.

Conclusions:

• Local monitoring is more efficient and effective for detecting contagious processes in large-scale networks.
• The proposed method provides an advance warning system for viral outbreaks.
• Applicable to monitoring various contagious phenomena across global-scale networks.