A website’s popularity proceeds like a snowball: The more people click on it, the more likely it is to rise to the top of a search, sending an ever higher stream of hyperlinks and clicks its way.
In a paper in Physical Review Letters, Jacob Ratkiewicz and colleagues at Indiana University and the Institute for Scientific Interchange in Italy study what indicators might best track how online popularity shifts over time. They looked at two large networks: Wikipedia pages and web sites in Chile. Within each network, they tracked changes⎯or “bursts”⎯in the number of hyperlinks to each of the sites and the amount of traffic this site received over time.
Network models that only assume “the more you have, the more you’ll get” fail to capture the distribution of bursts. The reason: these models don’t take into account external influences on people’s choices, such as new interest in an actress’ website after she wins a major award. Instead, Ratkiewicz et al. propose a “rank shift model” where, at each time interval, a web page randomly receives a new rank in overall popularity that is somewhere between where it currently is and the top (most viewed) position.
The authors found that their model compared well with the dynamics of bursts in Wikipedia pages in 2003⎯a data set large enough ( web pages) that they could see a realistic distribution of bursts but small enough that they could run a simulation. – Jessica Thomas