As information continues to grow at a very fast pace, our ability to access thisinformation effectively does not, and we are often realize how harder is getting tolocate an object quickly and easily. The so-called personalization technology is oneof the best solutions to this information overload problem: by automatically learningthe user profile, personalized information services have the potential to offer users amore proactive and intelligent form of information access that is designed to assistus in finding interesting objects. Recommender systems, which have emerged as asolution to minimize the problem of information overload, provide us withrecommendations of content suited to our needs. In order to providerecommendations as close as possible to a user?s taste, personalized recommendersystems require accurate user models of characteristics, preferences and needs.Collaborative filtering is a widely accepted technique to provide recommendationsbased on ratings of similar users, But it suffers from several issues like data sparsityand cold start. In one-class collaborative filtering, a special type of collaborativefiltering methods that aims to deal with datasets that lack counter-examples, thechallenge is even greater, since these datasets are even sparser. In this thesis, we present a series of experiments conducted on a real-life customer purchase databasefrom a major Turkish E-Commerce site. The sparsity problem is handled by the useof content-based technique combined with TFIDF weights, memory basedcollaborative filtering combined with different similarity measures and also hybridsapproaches, and also model based collaborative filtering with the use of SingularValue Decomposition (SVD). Our study showed that the binary similarity measureand SVD outperform conventional measures in this OCCF dataset. |