Issue No.05 - September/October (2007 vol.22)
Daniel Zeng , University of Arizona
Hsinchun Chen , University of Arizona
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/MIS.2007.80
An evaluation of six recommendation algorithms on e-commerce-related data sets is an initial step toward a metalevel guideline for choosing the best algorithm for a given application with certain data characteristics. Two of the evaluated algorithms are the popular user-based and item-based correlation/similarity algorithms. The other four algorithms attempt to meet the challenge of data sparsity through dimensionality reduction, generative models, spreading activation, or link analysis. Initial experimental comparisons indicate that the link-analysis algorithm achieves the best overall performance across several e-commerce data sets.
recommender systems, collaborative filtering, algorithm design and evaluation, e-commerce
Daniel Zeng, Hsinchun Chen, "A Comparison of Collaborative-Filtering Recommendation Algorithms for E-commerce", IEEE Intelligent Systems, vol.22, no. 5, pp. 68-78, September/October 2007, doi:10.1109/MIS.2007.80