TWITTER and PAGE RANK?
https://people.scs.carleton.ca/~maheshwa/courses/3801/Projects17/Popularity-Twitter-Report.pdf
The paper is about the potential of the generic Page Rank algorithm being applied to multiple other systems that involve ranking importance of elements. Specifically, this paper focused on Twitter and how PageRank can be applied to the social network graph of its follower relations. To conduct his research the author used a snapshot of the Twitter dataset as it appeared in 2010. The social network graph of Twitter at this time consisted of 41,652,230 users, with 1,468,365,182 follower relations. Then after making the data usable he applied to the Page Rank rule on it.
Before applying page rank, it is important to note that the PageRank algorithm gives a bigger rank to pages that have large indegrees. Twitter, on the other hand, looks at follower relations and looks for accounts with large outdegrees and to determine popularity. However, when applying the algorithm as-is on the Twitter dataset, we are doing a Reverse PageRank. Nodes with a lower score are the nodes that are more popular. This can be solved by using this Reverse PageRank as an account’s popularity by treating it in the same manner as golf scores: the lower the score, the more popular the account, and we order by ascending score rather than descending. “FollowerRank” of each node to use as a basis for comparison. This calculation is simple; a nodes FollowerRank is its outdegree divided by the total number of edges in the graph “(the percentage of follower relations the node has out of all follower relations)”.
After running both algorithms, there were some discrepancies in the data. For example, Obama has a very different rank in both. This is because while he does have a lot of followers his ratio isn’t as good as, for instance, Oprah. Also, the presence of bot accounts seems to have screwed with the data as well. PageRank does not seem particularly suited to evaluating a Twitter ac- count’s popularity, at least not without some modifications.
It relates to the topic of page rank we learned in class. We learned it the normal way that didn’t require a reverse. But cases where it does work would be in google who uses page rank to determine importance of pages. PageRank is a strong asset that connects search, advertising, recommendation and reputation systems. “The merit of PageRank comes from its power in evaluating network measures in a connected system”.