| Abstract |
The World-Wide-Web has emerged during the last decade as one of the most prominent research fields. However, its size, heterogeneity and complexity to a large extent overcome our ability to efficiently manipulate data using traditional techniques. In order to cope with these characteristics several Web applications require intelligent tools that may help to extract the proper information relevant to the users requests. In this thesis we report on the algorithmic aspects of recommendation technologies, which refer to algorithms and systems that have been developed to help users find items that may be of their interest from a variety of available items. Collaborative Filtering (CF), the prevalent method for providing recommendations, has been successfully adopted by research and industrial applications. However, its applicability is limited due to the sparsity and the scalability problems. Sparsity refers to a situation that transactional data are lacking or are insufficient, while scalability refers to the expensive computations required by CF. For addressing the scalability problem we propose a method of Incremental CF (ICF) that is based on incremental updates of user-to-user similarities. Our ICF algorithm (i) is not based on any approximation method, thus it gives the potential for high-quality recommendations formulation, and (ii) provides recommendations orders of magnitude faster than classic CF and thus, is suitable for online application. To provide high-quality recommendations even when data are sparse, we propose a method for alleviating sparsity using trust inferences. Trust inferences are transitive associations between users in the context of an underlying social network and are valuable sources of additional information that help dealing with the sparsity and the cold-start problems. Our experimental evaluation indicates that our method of trust inferences significantly improves the quality performance of the classic CF method. Finally, we provide a roadmap for future research directions that extend recommendation technologies to more complex types of applications and identify various research opportunities for developing them.
|
Επιλύοντας τα Προβλήματα Σποραδικότητας και Κλιμακοσημότητας των Αλγορίθμων Συστάσεων
