Recommender Systems Review of Types, Techniques, and Applications

Recommender Systems Review of Types, Techniques, and Applications

George A. Sielis (University of Cyprus, Cyprus), Aimilia Tzanavari (University of Nicosia, Cyprus & Cyprus University of Technology, Cyprus), and George A. Papadopoulos (University of Cyprus, Cyprus)
Copyright: © 2015 |Pages: 11
DOI: 10.4018/978-1-4666-5888-2.ch714
OnDemand PDF Download:
No Current Special Offers


Recommender or recommendation systems are software tools that make useful suggestions to users, by taking into account their profile, preferences and/or actions during interaction with an application or website. They are usually personalized and can refer to items to buy, people to connect to or books/ articles to read. Recommender Systems (RS) aim at helping users with their interaction by bringing to surface the information that is relevant to them, their needs, or their tasks. This article's objective is to present a review of the different types of RS, the techniques and methods used for building such systems, the algorithms used to generate the recommendations and how these systems can be evaluated. Finally, a number of topics are discussed as envisioned future research directions.
Chapter Preview

Recommendation Filtering Techniques/Algorithms

In general RS refer to the production of recommendations to be presented to a user, where these recommendations are useful to the user for the accomplishment of a task. This task might be related to navigation to web pages that interest the user, finding items to buy, explore learning resources or find people to socialize with or collaborate with. The types of recommendations are usually domain and task dependent and consequently context dependent. Specifically, recommendations systems can be identified as content-based, collaborative, or hybrid, depending on the basis of the filtering technique they use. In this section we describe the high level RS architecture and describe the several recommendation techniques that can be used for the development of a RS.

Key Terms in this Chapter

Similarity Distance: The distance between preference attributes within a geometric vector based space.

Recommendation Filtering Techniques: Techniques that are used to filter the data in order to make the data compatible to the standard RS model which includes the three main concepts user-items-ratings. Collaborative Filtering, Content Based Filtering and Hybrid filtering are the most known techniques. The selection of a filtering technique depends on the type of data which will be filtered.

Recommendation Frameworks: Set of layered functions that are available to software developers as tools for the development of recommendation systems.

Multi Criteria Recommender Systems: Recommender Systems that incorporate preference information upon multiple criteria.

Recommender systems: Software systems designed to filter data information and based on this predict the ratings or preferences for particular users in relation to items.

Recommendation Systems Evaluation Metrics: Metrics used for the evaluation of recommendation systems algorithms. Metrics can be defined based on pure mathematic formulas but also based on objective opinions of end users.

Complete Chapter List

Search this Book: