How Facebook Reads User Data and Desires

How Facebook Reads User Data and Desires – Facebook reveals how to guess the feed seen by readers, including using a machine learning (ML) algorithm system and predictors. This feature is claimed to support the News Feed with many layers.

According to the official page, Facebook processes trillions of posts that are presented to its more than 2 billion users. It establishes posts and thousands of signals to select posts that are relevant to its users.

It says when a user opens Facebook, the selection process takes place behind the scenes within seconds to create a feed of posts that the user opens.

Once that's running, there are several layers of ML models and algorithms applied to predict meaningful and relevant content for each user.

As a user goes through a number of stages, an algorithmic system narrows down the thousands of Feed candidates to the few hundred that appear in someone's News Feed at any given time.

In simple terms, the system determines which posts appear in a user's News Feed, and in the order in which the user is most interested. Reading this algorithm is based on several factors, including what and who users follow, who likes and with whom users interact.

For example, a user named Juan sees a photo post of a friend's rooster and he sees a video post of another friend's morning run. But Juan reposted an article on how to see the Milky Way at night.

All of this content is considered relevant or interesting by Juan, so Facebook predicts that the News Feed will have the highest value for Juan. In mathematical terms, Facebook defines an objective function for Juan and performs objective optimization.

Furthermore, Facebook also processes the characteristics of posts and photo posts tagged by its users. To rank more than a thousand user posts per day to make the process efficient.

It manages the data in multiple steps, which are efficiently organized to quickly limit the amount of computing resources required.

Next, the system rates posts on various factors, such as how well a post matches a user to interact with other users.

To count more than 1,000 posts and share them with billions of users at the same time, Facebook uses a system in the form of a post-reading parallel engine called Predictor.

Predictor is an intelligence engine for combining and narrowing interesting posts into approximately 500 posts.

Facebook runs contextual features such as the diversity of content types that are of interest to users. All of these steps happen in a matter of seconds, and the user has a pre-scored News Feed ready to be explored.


Back to top button