Descripción
|
|
---|---|
Recommender systems are offered as collaborative services, normally within the framework of the Web 2.0, and their use is very generalized, especially in the areas of e-commerce and e- learning. In order to generate recommendations, recommender systems must include a data base that contains the rated values of all the users with regards to the possible items involved. This could involve the creation of a database that is of a considerable size (for example, that of amazon.com). In this paper are presented the equations that permit the determination of which are the users and/or items whose contribution to the results of the system are the least beneficial, with the aim of being able to discard them during the recommendation process carried out during the collaborative filtering phase. With the aim of demonstrating how well these equations work, we have designed and carried out 98 comparative experiments based on the MovieLens database obtaining results that improve the performance of the recommender system at the same time as increasing their accuracy levels. The proposed method permits the size of the data bases of the recommender systems to be reduced, improves the accuracy of the results, rejects the items that contribute little or nothing to the quality of the recommendations and detects the least representative users of the system. | |
Internacional
|
Si |
Nombre congreso
|
International Conference on Internet Computing (ICOMP) |
Tipo de participación
|
960 |
Lugar del congreso
|
|
Revisores
|
Si |
ISBN o ISSN
|
1-60132-11-4 |
DOI
|
|
Fecha inicio congreso
|
13/07/2009 |
Fecha fin congreso
|
16/07/2009 |
Desde la página
|
196 |
Hasta la página
|
202 |
Título de las actas
|
Proceedings of the 2009 International Conference on Internet Computing, ICOMP 2009 |