Compound Classification Models for Recommender Systems
This event took place on Friday 19 May 2006 at 15:00
Prof. Dr. Dr. Lars Schmidt-Thieme University of Freiburg
Recommender systems recommend products to customers based on ratings or past customer behavior. Without any information about attributes of the products or customers involved, the problem has been tackled most successfully by a nearest neighbor method called collaborative filtering in the context, while additional efforts invested in building classification models did not pay off and did not increase the quality. Therefore, classification methods have mainly been used in conjunction with product or customer attributes.
Starting from a view on the plain recommendation task without attributes as a multi-class classification problem, we investigate two particularities, its autocorrelation structure as well as the absence of re-occurring items (repeat buying). We adapt the standard generic reductions 1-vs-rest and 1-vs-1 of multi-class problems to a set of binary classification problems to these particularities and thereby provide a generic compound classifier for recommender systems. We evaluate a particular specialization thereof using linear support vector machines as member classifiers on MovieLens data and show that it outperforms state-of-the-art methods, i.e., item-based collaborative filtering.
This event took place on Friday 19 May 2006 at 15:00
Prof. Dr. Dr. Lars Schmidt-Thieme University of Freiburg
Recommender systems recommend products to customers based on ratings or past customer behavior. Without any information about attributes of the products or customers involved, the problem has been tackled most successfully by a nearest neighbor method called collaborative filtering in the context, while additional efforts invested in building classification models did not pay off and did not increase the quality. Therefore, classification methods have mainly been used in conjunction with product or customer attributes.
Starting from a view on the plain recommendation task without attributes as a multi-class classification problem, we investigate two particularities, its autocorrelation structure as well as the absence of re-occurring items (repeat buying). We adapt the standard generic reductions 1-vs-rest and 1-vs-1 of multi-class problems to a set of binary classification problems to these particularities and thereby provide a generic compound classifier for recommender systems. We evaluate a particular specialization thereof using linear support vector machines as member classifiers on MovieLens data and show that it outperforms state-of-the-art methods, i.e., item-based collaborative filtering.
Future Internet
KnowledgeManagementMultimedia &
Information SystemsNarrative
HypermediaNew Media SystemsSemantic Web &
Knowledge ServicesSocial Software
Social Software is...
Interacting with other people not only forms the core of human social and psychological experience, but also lies at the centre of what makes the internet such a rich, powerful and exciting collection of knowledge media. We are especially interested in what happens when such interactions take place on a very large scale -- not only because we work regularly with tens of thousands of distance learners at the Open University, but also because it is evident that being part of a crowd in real life possesses a certain 'buzz' of its own, and poses a natural challenge. Different nuances emerge in different user contexts, so we choose to investigate the contexts of work, learning and play to better understand the trade-offs involved in designing effective large-scale social software for multiple purposes.
List all Social Software Projects
Check out these Hot Social Software Technologies:
List all Social Software Technologies
