The main idea of collaborative recommendation approaches is to exploit information about the past behavior or the opinions of an existing user community for predicting which items the current user of the system will most probably like or be interested in. These types of systems are in widespread industrial use today, in particular as a tool in online retail sites to customize the content to the needs of a particular customer and to thereby promote additional items and increase sales.

From a research perspective, these types of systems have been explored for many years, and their advantages, their performance, and their limitations are nowadays well understood. Over the years, various algorithms and techniques have been proposed and successfully evaluated on real-world and artificial test data.

Pure collaborative approaches take a matrix of given user-item ratings as the only input and typically produce the following types of output: (a) a (numerical) prediction indicating to what degree the current user will like or dislike a certain item and (b) a list of n recommended items. Such a top-N list should, of course, not contain items that the current user has already bought.

User-based nearest neighbor recommendation

The first approach we discuss here is also one of the earliest methods, called user-based nearest neighbor recommendation. The main idea is simply as follows: given a ratings database and the ID of the current (active) user as an input, identify other users (sometimes referred to as peer users or nearest neighbors) that had similar preferences to those of the active user in the past.