Dynamic Recommender Systems
Recommender Systems gain popularity in recent days. Numerous companies recognized the potential of recommender systems and use them with success. The most remarkable examples are Amazon, Netflix, Youtube, etc. The goal of this project is the development of recommender systems that are able to learn user preferences from fast and dynamic data streams. The main challenges are constant changes of the environment and capturing of users' evolving preferences. The most successful methods in recommender systems are based on matrix factorization. Those methods reveal high accuracy also on sparse data. However, the most of them work on static datasets, which makes them inapplicable in real world scenarios. One of the goals of the project is to make those methods incremental and adaptive to changes over time. Further challenges are high efficiency requirements and constantly changing data space.
- Predicting the Performance of Collaborative Filtering Algorithms. Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14), 38:1--38:6, ACM, New York, NY, USA, 2014. URL