Υλοποίηση και συγκριτική αξιολόγηση μεθόδων συνεργατικής διήθησης
Ηρακλειδής, Γεώργιος Β.
Recommender systems try to simulate people preferences for the purpose of estimate how much a user will be interested in some objects, information or services and to place a suggestion for items an individual should buy or at least examine. These systems is being used constantly in the web and have become basic components of electronic commerce and information retrieval, making suggestions that filter large information spaces so as to every user could be shown items that suits well his needs and interests. The huge growth of available information and items in the internet as well as the constantly increasing number of visitors of web sites in the recent years poses some challenges for recommender systems. These are the production of accurate recommendations, the calculation of large number of recommendations in a very small time period and the successful completion of a very big proportion of the requests that the system accepts. Various techniques like collaborative filtering, content based and demographic as well as some combinations of them have been used to create recommendations. This postgraduate thesis surveys a number of collaborative filtering algorithms analyzing the steps for creating such a system and compares which methods provide better results in the challenges that a recommender system encounters.