Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Supervisors info:
Ιωάννης Ιωαννίδης, Καθηγητής, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Μαριαλένα Κυριακίδη, Υποψήφια Διδάκτορας, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Original Title:
Χρονικές συστάσεις με βραχυπρόθεσμες και μακροπρόθεσμες προτιμήσεις
Translated title:
Temporal recommendations with long-term and short-term preferences
Summary:
Time is an important factor when making recommendations and accurately capturing user preferences over time is a great practical challenge in recommender systems.
Collaborative Filtering (CF) algorithms, used to build web-based recommender systems, are often evaluated in terms of how accurately they predict user ratings and many current
evaluation techniques disregard the fact that users continue to rate items over time and change their preferences due to different external events. User behavior can often be determined by individual’s long-term and short-term preferences.
To address these challenges, the first method we used was the session-based Temporal Graph (STG) which simultaneously models users’ long-term and short-term preferences over time. Based on the STG model framework, we used the recommendation algorithm Injected Preference Fusion (IPF).
For the second part we tried a different approach with the collaborative filtering, by using Principal Component Analysis (PCA) and hierarchical clustering, to group similar users and making user-based recommendations.
Finally, we evaluate the effectiveness of the methods using Yelp dataset. Based on business reviews and making recommendations, we prove that the STG method presents more accurate results than the PCA method.
Main subject category:
Science
Keywords:
temporal recommendation system, collaborative filtering, graph, clustering, top-n list
