The People You May Know recommender system accounts for more than 50% of the connections on LinkedIn, and is an excellent example of how Machine Learning is found in everyday life. It churns through hundreds of terabytes of data every day, processing hundreds of billions of potential connections.
Here are some "signals" the People You May Know recommender engine uses to compute the likelihood that people will connect:
- A single shared connection - If George and Jerry are friends, George may also know other people (like Elaine) that are connected to Jerry.
- Multiple shared connections - If Jerry knows George and Elaine, and George and Elaine are both friends with Newman, it's probable that Jerry also knows Newman.
- Shared companies - If George and Elaine worked at the same company during date ranges that overlap, it increases the probability that they know each other.
- Shared groups - George is even more likely to know friends of Jerry if they share any of his organizations or interests.
- Location
- Age Difference
- How big an employer is
- How "social" a company is
- The timeframe within which someone joins and leaves a company
The People You May Know recommender system takes into account active user feedback and implied user feedback (whether they interact with a recommendation or ignore it). The system actually "discounts" the value of a recommendation if it has been ignored over time and re-prioritizes the recommendations that are shown.
See more examples of Machine Learning in our Everyday Encounters blog series >>
