Collaborative filtering

In collaborative filtering ( collaborative filtering ) are evaluated by groups of users to infer the interests of individuals behavior. This is a form of data mining, which makes an explicit user input superfluous.

Target

The use of collaborative filtering is done mostly for very large amounts of data. Collaborative filtering is used for a variety of areas such as in the financial services sector for the integration of financial sources, or in applications in e-commerce and Web 2.0. This article deals with the collaborative filtering for user data, although some methods and approaches can be transferred to other areas.

The aim of the method is an automatic prediction (filtering ) of user interests. For this purpose, information about the behavior and preferences of many users as possible are collected ( collaboration). The underlying assumption of collaborative filtering is that someone who is looking for something (eg a TV broadcast ) was interested in the past, will also be interested in the future of this. Through collaborative filtering can be made for the TV, a prediction which telecast could please a crowd. The output would be a list of possible favorite television broadcasts. It should be noted that prediction is made for each viewer individually. The data basis for the prediction is collected by the community of users. Here, the collaborative filtering of simpler methods, in which a non-specific mean value is calculated is different.

A specific problem of collaborative filtering is its latency: A new user enters the system with an empty user profile. Since its interests are not yet known, it can get at the beginning of any meaningful recommendations. The same applies to new entrants to the system elements ( eg products in an online store ). They have no quantifiable similarity with other elements and can not therefore be meaningfully recommended. It is in collaborative filtering ie by learning systems, and thus a form of artificial intelligence.

Methodology

Collaborative filtering mostly runs in two steps.

Alternatively, there is the item-based collaborative filtering, which was known by Amazon.com ( "This could be interesting. " ) And was first introduced by Vucetic and Obradovic in 2000.

Other forms of Collaborative filtering can be based on implicit observation of user behavior. In these forms of filtering the behavior of the individual user is with the behavior of all other users compared ( What music have you heard? What products do you have? ). This data can be used for the future behavior of the user's predictions. It does not make sense to offer a user a particular piece of music, when he has made it clear by his behavior that he has it already. Likewise, it does not make sense to offer a more user - Paris travel guide, if he already has a guide for this city.

In today's information age, these and similar technologies turn out to be extremely helpful for product selection, just when certain product groups (eg music, movies, books, news, web pages ) have become so large that individual persons can not see the entire range.

Application

In commercial systems

Commercial websites that use collaborative filtering:

  • Amazon
  • Amie Street
  • Barilliance
  • Barnes and Noble
  • Baynote
  • Choice stream
  • Collarity
  • Digg.com
  • EBay
  • Google News
  • Gravity R & D
  • Half.ebay.com
  • Heeii
  • Hollywood Video
  • Hulu
  • ILike - Music
  • Internet Movie Database - Movies
  • ITunes - Music
  • Last.fm - Music
  • LibraryThing - Books
  • Loomia - Software as a Service provider of Recommendation technologies
  • MusicMatch
  • MyStrands - developer of social recommendation technologies
  • Netflix
  • Simania - book recommendation page
  • Strands - Strands uses and advertises its own recommendation engine for social networks and eCommerce
  • StumbleUpon - website
  • Threadless - T-Shirt
  • TiVo
  • Yelp
  • Ramkol - Sophisticated recommendation for local search in Israel

In non-commercial systems

  • AmphetaRate - RSS Articles
  • Everyone's a Critic - Movies
  • GiveALink.org - websites
  • Gnomoradio - Music
  • Movie Lens - Movies
  • Rate Your Music Music
Data-Mining Collaborationism Barnes & Noble Yelp, Inc. Integrated Authority File
482693
de