What are the major machine learning theories that maybe used by Twitter for suggesting followers?
Three different approaches come to my mind
- decisions based solely on the following / follower lists. If a lot of people you are following, follow a particular person, the chance is high you might be interested in this persons tweet.
- using links and hashtags. Assuming you link very often to specific websites or use specific links you might be interested in people doing the same.
- doing some kind of document clustering approaches on the tweets, to figure out who's writing similar things and suggest him or her.
The problem with recommender systems to me is that most machine learning algorithms are just using some kind of similarity measure to give suggestions.
Check out "recommender systems" and "document clustering" as search keywords to get some more ideas.