Collaborative filtering prediction mechanism and recommendation system

Introduction :

Recommendation systems are at the heart of our web activity. The objective is to produce a personalized list of suggestions (products, links to click, …) for a “customer” (visitor of a site, potential buyer of a product, …) in relation with his concerns and expectations.

• Interest for the “customer” : (quick) orientation towards the most relevant elements, reduce the search time, improve the user experience…
• Interest for the “seller” : more sales, more clicks, keeping the customer’s interest, making him stay longer, directing him to certain products, building loyalty…

Faced with the quantity and speed of new information, the development of information systems to target the answers provided to users so that they are closer to their expectations and personal tastes has become an unavoidable necessity.

Collaborative filtering systems are among these information systems with certain particularities that make the difference.

The term collaborative filtering refers to techniques that use the known tastes of a group of users to predict the unknown preference of a new user.

This paper describes a basic collaborative filtering mechanism that allows users to :

• the discovery of new items (products, articles, etc.), through the automation of the natural recommendation process,