Recommender system

Recommender system (RS) emerged to provide solutions for the information overload; today many web sites are being created and changed every day and the amount of information has been increased dramatically creating vast amounts of content in the internet network. RS aims to provide users with the items (movies, music, books, news, web pages) best match their individual preference.

Today many web based RS Application being developed for various domains, and many on-line stores (e.g. Amazon, MediaUnbound,Netflix, MoodLogic, CDNOw, SongExplorer, etc) recommend items that meet their customers’ preferences and hence substantially increase their sales.

Algorithms for RS can be divided into two types, model based and memory based. Model-based algorithms use the collection of ratings to learn a model, which is then used to predict the ratings, similar to the method used in data mining for discovering patterns from observations. In these algorithms we only store the model, rather than the ratings and for each request, the model is consulted for prediction. Examples for model based algorithms are clustering, Bayesian model, and k-means.

In the other side, Memory based algorithms use heuristics that predict new items based on the entire collection of previously rated items by other similar users. The predicted item is usually computed as an aggregate of the ratings of some other users (k similar users) for the same item.

Content-Based approach is the first approach to be proposed for RS, it comes from the information retrieval field. In this approach the user will be recommended items similar to the ones the user preferred in the past, or to the areas of interests of the user (Balabanovic&Shoham, 1997a;Clypool et al .,1999).

Today Web 2.0 offered new opportunities for RS, social networks (e.g., Myspace, Facebook) and virtual communities are its main lineament which might have a huge contribution for RS. Recent researches detected this potential and start investigating the impact of social relations in RS, to date most of this researches focus mainly on trust, while the behavioral theory suggest other social relations that impact people’s advice taking (e.g., communication frequency, reputation), these social relations were neglected. Users communicate via E-mail application (e.g., Gmail: Google mail service) and produce interaction social tie, users can contact friends and search for new friends (e.g., Facebook) producing friendship social ties; others may participate in online auctions in C2C trade environment (e.g., eBay.com) producing reputation ties. These are a few example of different platform which enable members to establish different social ties with each others, such these social ties are more available today than before, and we believe that such information regarding users’ relationships could potentially be exploited for recommender system.

---

Possibly related posts: (automatically generated)

• Explaining recommendations
• Our Project assigment!