direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Logo der TU Berlin

Graph-based Preference Collecting Interface

Graph-based Preference Collecting Interface


Masterarbeit (ggfs. auch Bachelor)


Recommender Systems support users to overcome the information overload caused by huge item collection. Collaborative Filtering has established as state-of-the-art technique. Preferences assigned by users towards items play a key role in the process of generating recommendations. Typically, recommender systems utilise a fixed set of ratings to let users express their preferences towards items. Having a fixed set of ratings entails a fuzzy grading issue. Users cannot express fine-grained differences in their preferences. Additionally, evaluating recommender systems becomes a hard task. Usually, recommender systems are evaluated either by how accurately preferences can be predicted or how well items can be ranked according to their actual preferences. The latter evaluation suffers from fuzzy rankings due to fixed ratings. Letting users express their preferences as preference graph ought to overcome those issues.


1. The thesis shall include a literature review covering essential works. Relevant areas of research include recommender systems (motivation, algorithms, evaluation, etc.), user interface design and graph theory.
2. The thesis shall develop a concept for collecting user preferences in form of a preference graph.
3. The student ought to implement a user interface with whom graph-based preferences can be collected.
4. The thesis shall include an evaluation of how well the implemented user interface performs. The evaulation ought to cover aspects such as scalability, stability, and consistency.
5. The studen must motivate significant design choices as well as introduce alternative concepts.


Language: English


Betreuer:  Benjamin Kille
Email:   Benjamin.Killedai-labor.de