| Περίληψη: | This chapter discusses the technology and applications of recommender systems. It presents the fundamental architectures and data analysis techniques that are currently used. The techniques include the content-based, the collaborative filtering, the knowledge-based and the hybrid ones. The chapter continues with the description of the major problems that restrict recommender systems success. The problems list includes the sparse data problem. There is often the case where ratings are sparse and not available for many service and product features, also not all users declare disclose their needs. The data Sparsity refers to the case where not enough product ratings or user behavioral data is available, thus making recommendations difficult or less useful. Next, the chapter discusses the cold-start problem, which implies that in the early stages of using a recommender system, the system lacks important information of user ratings, so it is difficult if not possible to suggest a user services without knowing almost anything about the users’ needs and priorities. The third major problem discussed is that of fraud. There are cases where users are trying to manipulate the recommender systems in favor of certain services, called push attacks, or in order to downgrade other, called nuke attacks. Finally, this chapter discusses recommender systems evaluation methods, as well as application areas and real life examples.
|
|---|
