The concept drift problem, which arises from the evolution over time of real-world concepts, has long been investigated. Users’ interests and preferences are among the most significant examples of it. This issue can heavily affect personalized systems, whose accuracy can be decreased as users' interests evolve over time. Taking into account the temporal dynamics of users' interests can, therefore, bring significant benefits to adaptive systems.

In this talk, we will analyze some user modeling approaches proposed in the research literature, which exploit different techniques to take into account the temporal dynamics of users’ interests and preferences. Specifically, we will consider classical approaches based on Vector Space Model and Matrix Factorization, approaches based on Signal Processing, and approaches based on Recurrent Neural Networks.

We will also discuss the results of a comparative analysis performed by applying these approaches for the realization of a social recommender system, that is, a system capable of providing the target user with suggestions of social media (e.g., Twitter) users of her possible interest. This analysis was conducted on a dataset of over 2,700,000 tweets collected by monitoring a sample of over 1,600 users for a full year.

Bio: Giuseppe Sansonetti is an assistant professor with tenure track at the Artificial Intelligence Laboratory of the Department of Engineering, Roma Tre University, where he is currently teaching the “Intelligent Systems on the Internet” and “Artificial Intelligence and Machine Learning” courses in Computer Science. He received his Master's Degree (summa cum laude) in Electronic Engineering and Pd.D. in Computer Science. From 2003 to 2005 he pursued research at the University of California, Santa Barbara (UCSB), and Oregon State University (OSU). He also worked as Research Fellow at the Italian Institute for Nuclear Physics (INFN) and the National Interuniversity Consortium of Materials Science and Technology (INSTM). He has been involved in National and International research projects regarding Internet technologies. He has been a reviewer of well-established International Journals and Conferences. His current research interests include user modeling, recommender systems, case-based reasoning, and computer vision.