A review on matrix factorization techniques in recommender systems
Authors:
- 1 April 2017
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 269-274
Abstract
Growth of the Internet and web applications has led to vast amount of information over web. Information filtering systems such as Recommenders have become potential tools to deal with such plethora of information, help users select and provide relevant information. Collaborative Filtering is the popular approach to recommendation systems. Collaborative Filtering works on the fact that users with similar behavior will have similar interests in future, and using this notion collaborative filtering recommends items to user. However, the sparseness in data and high dimensionality has become a challenge. To resolve such issues, model based, matrix factorization techniques have well emerged. These techniques have evolved from using simple user-item rating information to auxiliary information such as time and trust. In this paper, we present a comprehensive review on such matrix factorization techniques and their usage in recommenders.Keywords
This publication has 21 references indexed in Scilit:
- Recommender Systems: Sources of Knowledge and Evaluation Metrics2013
- Distributed Matrix Completion2012
- Comparison of collaborative filtering algorithmsACM Transactions on the Web, 2011
- Factorization meets the neighborhood2008
- Scalable collaborative filtering using cluster-based smoothing2005
- Slope One Predictors for Online Rating-Based Collaborative Filtering2005
- Are Loss Functions All the Same?Neural Computation, 2004
- Amazon.com recommendations: item-to-item collaborative filteringIEEE Internet Computing, 2003
- Item-based collaborative filtering recommendation algorithms2001
- Using collaborative filtering to weave an information tapestryCommunications of the ACM, 1992