Recommender Systems are everywhere. From Amazon recommending books or other items with “Customers who bought this item also bought ..” to Netflix recommending movies based on the user’s viewing habits, we see recommender systems in action on a daily basis.
Recommender systems have come a long way from the early days of Amazon (e.g., Amazon’s seminal paper) and Netflix’s million dollar prize in 2006. Although consumer web seems to have paved the way, recommendation systems are now common in enterprise applications as well.
Last summer, I worked as a Data Science Intern in one of the leading startups that provides TXP (Talent Experience Platform) solutions – one that helps Fortune 50 companies like Novartis with Talent Mobility solutions. In its core “Spotify of Learning” product that helps users to navigate information overload issue (e.g., Sally wants to move from Software Engineering to Product Management. What learning pathways and contents should she follow to be able to successfully do that), EdCast uses many of the recommender system techniques, including collaborative filtering.
Another friend of mine interned at a startup that’s building a CRM (Customer Relationship Management) system targeted for the automotive industry. The system helps consumers make service appointments for maintenance as well as repairs for their cars. A recommendation engine is used during the appointment making process. For example, let’s assume that Toyota Corolla and Honda Civic are two “similar” types of cars. When a user makes an appointment for a Honda car, she would get a recommendation for maintenance services not only based on other Honda Civic cars but also other Toyota cars in the platform.
As I am digging more into the algorithms behind the recommendation engines, I have been reading about content based filtering vs collaborative filtering and matrix factorization. Here are some of the sites I have been following:
The Data Chat 
Leave a comment