Skip to main content



The EdgeRank Algorithm: Optimizing Your News Feed

Facebook is an incredible marketing machine. Every component of the popular networking site and every update draws users to increase their use of Facebook with a clear profitable motivation. Certain features are solely to hold your attention to the site such as the real time chat system, leading to greater ad viewership and general usage. The updated layout with all essential information posted to the very top of your profile (current location, birthday, languages, career, etc.) encourages users to hand over more facts for Facebook’s targeted marketing. All features and updates are deliberate and specially designed to benefit the money-making marketing powerhouse and the news feed is no exception.

What makes the news feed interesting? Facebook uses many hidden metrics to prioritize information, mathematically determining which stories are important and which stories are irrelevant. This is done by observing multiple networks formed by the site’s acitivity. We’re all aware that the mutual relationship of friendship between users forms a network, but there is also a hidden sub-network. Facebook treats every input created by users as an object (photos, statuses, etc.). When an object is created by a user, this forms another node in the network. This node from the creator is a is connected to all other users with a specific weight that determines how important this input object is to all of your friends. The metric used is a function using your social network graph called EdgeRank.

EdgeRank inputs three factors to determine the weighting of the edge between an object and a viewer: an affinity score between the viewer and the creator, a coefficient associated with object type and number of linked objects, and a simple decaying factor as a function of time. The affinity score comes from the original graph of the social network where all users represent nodes and each edge represents friendship. The weight on each edge is derived from Facebook’s friendship score and represents the degree of friendship between two users. Every time two users interact, this weight is increased accordingly. Each object type also corresponds to a unique coefficient. This is because certain objects create higher ad viewership by being a source of “clicks”, leaving photos and videos at the very top of Facebook’s object type hierarchy. Associated object types also increase this weighting term for the original object, with comments associated with the activity being the most important. Comments are valuable because they require users to engage Facebook with more focus. Finally, a time decay function causes the object to disappear from your news feed over time. The importance of an activity object is the sum over all the edges leaving the object’s node, which includes users and associated objects.

Problems always arise in the study of social networks where weighted edges are necessary for optimizing the use of those networks. As we have seen before, methods of weighing edges are necessary for improving recommendation systems. The EdgeRank algorithm suggests a simple cumulative method for measuring friendships between users. Using this weighting system we can determine a threshold between strong and weak relationships and observe the Strong Triadic closure property. The EdgeRank algorithm is just another example of ways we input graphs and networks to enhance the user’s experience.

Reference: http://thenextweb.com/socialmedia/2011/05/09/everything-you-need-to-know-about-facebook%E2%80%99s-edgerank/

Comments

Leave a Reply

Blogging Calendar

September 2011
M T W T F S S
 1234
567891011
12131415161718
19202122232425
2627282930  

Archives