Mining User Behavior in Social Environments

The growth of the Web 2.0 has brought to a widespread use of social media systems and to an increasing number of active users. This phenomenon implies that each user interacts with too many users and is overwhelmed by a huge amount of content, leading to the well know “social interaction overload” problem. In order to address this problem several research communities study Social Recommender Systems, which are information filtering systems that operate in the social media domain and aim at suggesting to the users items that are supposed to be interesting for them. Social Recommender Systems usually filter content by exploiting the social graph or by mining the user content. Since the social domain is characterized by a continuous and quick growth of the the amount of content and users, both these approaches face some problems to produce accurate and up-to-date recommendations. This PhD thesis proposes some social recommendation approaches based on the mining of the user behavior, i.e., on the exploitation of the activity of the users in social environments, in order to produce accurate and up-to-date recommendations.