Recommender systems refer to a technology that is able to predict user responses to specific user options by analyzing big data. One is example of such a system offers news articles to online newspaper readers that are based on a prediction of reader interest. Another example is to offer customers of an online retailer suggestions about what products they might like to buy that are based on their history of product purchases or even only online product searches. A recommendation engine that is able to provide this functionality can be categorized in two different types that are described below.
The first category of recommender systems are content-based systems that examine the properties of the items recommended. One example is movie recommendation such as it is performed in companies like Netflix that rents out movies to customers. The content-based recommender will learn that a user watched many science fiction movies and in turn will recommend a movie classified in the movie database as having the science fiction genre. The content of a science fiction movie is in this system the key for recommendation to the user.
The second category of recommender systems is known as collaborative filtering and recommend items based on similarity measures between different users or items.. This means that items are recommended to a specific user that are those preferred by similar users. In this context the process of identifying similar users and recommending what similar users like is called collaborative filtering since it filters options. Examples of similarity measures are Jaccard distance or Cosine distance. Compared to the first category above that is using features of items to determine their similarity the second category focusses on the similarity of the user ratings for two items. These recommendation engines thus play a big role in online advertising that involves selecting the right items to advertise at an online store. Often this includes to find customers with similar behavior in order to recommend to buy things that similar customers have bought before.
Recommender systems details
The following video provides more pieces of information about the topic: