This is more an update than an article. For those who don't know my big project for the last three weeks has been Appaholic, a great utility for graphing the growth of Facebook applications. It's been on the front pages digg and Mashable and has been making rounds in the blogosphere (today Robert Scoble linked to it).
I've also been given a guest blogger position at Inside Facebook, a blog about Facebook stuff. I'm going to put most of my Facebook editorial over there for now and keep 20bits more focused on code and technology. I think I'm going to write an article about using eigenvector centrality to determine the influencers in a social network, for example.
Anyhow, this was just a head's up and an explanation for why posting has been slower than before. Cheers!
It seems like eigenvector centrality is a clean representation of pagerank. Is it safe to say more central nodes are influencers?
Pedram,
PageRank is a variant of eigenvector centrality but includes some probabilistic factors.
The idea behind eigenvector centrality is that the score of a node is proportional to the sum of the score of the nodes to which it is connected. That seems roughly equivalent to real-world influence. If I’m friends with ten Fortune 500 CEOs I’m probably have a large influence, even if they’re my only friends.
I’m sure there are papers out there about this that, like PageRank, add additional factors to take into account features of social networks that aren’t a priori apparent, but I haven’t read any.