Modeling Behaviors of Browsing and Buying for Alidata Discovery Using Joint Non-negative Matrix Factorization
Authors:
- Bin JuBJBin Ju
- Mincao YeMYMincao Ye
- Yuntao QianYQYuntao Qian
- Rong NiRNRong Ni
- Chenxi ZhuCZChenxi Zhu
- 1 November 2014
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 401, 114-118
Abstract
In this paper, we propose a novel recommender algorithm for the data competition launched by the Alibaba Group based on the intuition that a user's buying behaviors will be influenced by the user's browsing behaviors on the web, which means that the latent preferences that lie behind these two behaviors are consistent. We present a matrix factorization framework that fuses a user-item buying matrix with a user-item browsing matrix using joint nonnegative matrix factorization. This approach assumes that the two factorized coefficient matrices obtained from the browsing matrix and the buying matrix should be regularized toward a common consensus. The experimental results show that our algorithm outperforms other algorithms based only on a single matrix factorization.Keywords
This publication has 7 references indexed in Scilit:
- Recommender systemsPhysics Reports, 2012
- A Survey of Collaborative Filtering TechniquesAdvances in Artificial Intelligence, 2009
- Matrix Factorization Techniques for Recommender SystemsComputer, 2009
- Toward the next generation of recommender systems: a survey of the state-of-the-art and possible extensionsIEEE Transactions on Knowledge and Data Engineering, 2005
- Latent semantic models for collaborative filteringACM Transactions on Information Systems, 2004
- Amazon.com recommendations: item-to-item collaborative filteringIEEE Internet Computing, 2003
- Learning the parts of objects by non-negative matrix factorizationNature, 1999