Predicting the Performance of Collaborative Filtering Algorithms
P. Matuszyk, and M. Spiliopoulou. Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14), page 38:1--38:6. New York, NY, USA, ACM, (2014)
DOI: 10.1145/2611040.2611054
Abstract
Collaborative Filtering algorithms are widely used in recommendation engines, but their performance varies widely. How to predict whether collaborative filtering is appropriate for a specific recommendation environment without running the algorithm on the dataset, nor designing experiments? We propose a method that estimates the expected performance of CF algorithms by analysing only the dataset statistics. In particular, we introduce measures that quantify the dataset properties with respect to user co-ratings, and we show that these measures predict the performance of collaborative filtering on the dataset, when trained on a small number of benchmark datasets.
%0 Conference Paper
%1 Matuszyk:2014:PPC:2611040.2611054
%A Matuszyk, Pawel
%A Spiliopoulou, Myra
%B Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14)
%C New York, NY, USA
%D 2014
%I ACM
%K kmd recsys
%P 38:1--38:6
%R 10.1145/2611040.2611054
%T Predicting the Performance of Collaborative Filtering Algorithms
%U http://doi.acm.org/10.1145/2611040.2611054
%X Collaborative Filtering algorithms are widely used in recommendation engines, but their performance varies widely. How to predict whether collaborative filtering is appropriate for a specific recommendation environment without running the algorithm on the dataset, nor designing experiments? We propose a method that estimates the expected performance of CF algorithms by analysing only the dataset statistics. In particular, we introduce measures that quantify the dataset properties with respect to user co-ratings, and we show that these measures predict the performance of collaborative filtering on the dataset, when trained on a small number of benchmark datasets.
%@ 978-1-4503-2538-7
@inproceedings{Matuszyk:2014:PPC:2611040.2611054,
abstract = {Collaborative Filtering algorithms are widely used in recommendation engines, but their performance varies widely. How to predict whether collaborative filtering is appropriate for a specific recommendation environment without running the algorithm on the dataset, nor designing experiments? We propose a method that estimates the expected performance of CF algorithms by analysing only the dataset statistics. In particular, we introduce measures that quantify the dataset properties with respect to user co-ratings, and we show that these measures predict the performance of collaborative filtering on the dataset, when trained on a small number of benchmark datasets.},
acmid = {2611054},
added-at = {2014-06-20T11:57:09.000+0200},
address = {New York, NY, USA},
articleno = {38},
author = {Matuszyk, Pawel and Spiliopoulou, Myra},
biburl = {https://www.bibsonomy.org/bibtex/2fa10c51811cc7b8ad8d7c41f0c0ad32f/kmd-ovgu},
booktitle = {Proceedings of the 4th International Conference on Web Intelligence, Mining and Semantics (WIMS14)},
doi = {10.1145/2611040.2611054},
interhash = {8d8d6b49d7b4ae2d7b8c80c32d1cd517},
intrahash = {fa10c51811cc7b8ad8d7c41f0c0ad32f},
isbn = {978-1-4503-2538-7},
keywords = {kmd recsys},
location = {Thessaloniki, Greece},
numpages = {6},
pages = {38:1--38:6},
publisher = {ACM},
series = {WIMS '14},
timestamp = {2014-09-04T16:11:36.000+0200},
title = {Predicting the Performance of Collaborative Filtering Algorithms},
url = {http://doi.acm.org/10.1145/2611040.2611054},
year = 2014
}