[thesis] The GReco Editor: Composing, Visualising, and Editing Group Recommender Systems

Schirmer, M. and Gross, T.

Master’s thesis, supervised by Tom Gross.

Abstract. Recommender systems provide a valuable amendment in many areas of human-computer interaction by facilitating the process of pre-selecting only those specific items out of a large collection that are relevant for individual users. They rely on ratings for items which enable them to estimate predictions for unrated items. Recommender systems are found in a variety of computing and information systems and support users during the interaction with these systems. Typical examples are online shopping services, movie recommendation services, or restaurant and travel information systems.

However, many of today’s hedonic activities are group activities. Regular recom- mender systems do not support groups of users and handle only single-user recommenda- tions. Group recommender systems on the other hand provide specialised algorithms for generating recommendations for groups. These systems aggregate the individual prefer- ences of all group members to estimate the prediction for the entire group. The concept, use and implementation of such group recommender systems present an interesting field of research. However, the configuration effort of these systems is very high. Tis limits end-users in their possibilities to explore and experiment with group recommender sys- tems. In this thesis, I present the concept and implementation of an advanced group recommender system that relies on two sophisticated group recommender algorithms that support individual group weight factors for the group members and incorporate explanations to illustrate the flow of events and decisions during the recommendation generation process.

Composing, visualising, and editing of the group recommender process is leveraged by a graphical editor application that makes use of advanced user interaction techniques in order to allow end-users, recommender systems developers, and researchers to explore, configure, and adjust the group recommender algorithms.