Multirelational Factorization Models
Factorization models are machine learning models that predict quantities
based on historical data, i.e., customer preferences, health risks,
etc. Faxtorization models specifically address problems where
interactions between objects should be predicted about which not many
data are known. They are very successfully employed in many areas,
foremost in business analytics where they are used to predict
customer behavior and try to recommend products to customers with
so-called recommender systems. ISMLL works on factorization models
for several years now. We have used them, e.g., in our paper on
sequential recommender systems that won the WWW 2010 best paper
award.
As applications typically can be described by not just a single
relation between objects, but by multiple such relations, e.g.,
relations such as customer-buys-product, customer-views-product,
product-has-property, etc., models that take into account multiple
such relations are one of the current very active research topics.
The project Multirelational Factorization Models will research
different such models and learning algorithms with both the aim
of using the additional information to build more accurate models
than models based on a single relation and the aim of a deeper
theoretical understanding why and when these models successfully
can be employed.
The project will run for three years from 2011 - 2014.
Basic Information
Sponsor: German Research Foundation
Contact
Publications
- Ernesto Diaz-Aviles, Lucas Drumond, Zeno Gantner, Lars Schmidt-Thieme, Wolfgang Nejdl (2012):
What is Happening Right Now ... That Interests Me? Online Topic Discovery and Recommendation in Twitter , Proceedings of the 21st ACM International Conference on Information and Knowledge Management (CIKM 2012) .
- Artus Krohn-Grimberghe, Lucas Drumond, Christoph Freudenthaler, Lars Schmidt-Thieme (2012):
Multi-Relational Matrix Factorization using Bayesian Personalized Ranking for Social Network Data , Proceedings of the Fifth ACM International Conference on Web Search and Data Mining . - Lucas Drumond, Steffen Rendle, Lars Schmidt-Thieme (2012):
Predicting RDF Triples in Incomplete Knowledge Bases with Tensor Factorization, in Proceedings of the 27th ACM International Symposium on Applied Computing, Riva del Garda, Italy. - Nguyen Thai-Nghe, Lucas Drumond, Tomáš Horváth, Lars Schmidt-Thieme (2011):
Multi-Relational Factorization Models for Predicting Student Performance, in KDD 2011 Workshop on Knowledge Discovery in Educational Data (KDDinED 2011). Held as part of the 17th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.