%0 Journal Article %1 Mclaughlin04 %A Mclaughlin, Matthew R. %A Herlocker, Jonathan L. %B SIGIR '04: Proceedings of the 27th annual international conference on Research and development in information retrieval %C New York, NY, USA %D 2004 %I ACM Press %K collaborative_filtering evaluation %P 329--336 %R 10.1145/1008992.1009050 %T A collaborative filtering algorithm and evaluation metric that accurately model the user experience %U http://dx.doi.org/10.1145/1008992.1009050 %@ 1581138814

@article{Mclaughlin04, added-at = {2009-06-22T17:28:38.000+0200}, address = {New York, NY, USA}, author = {Mclaughlin, Matthew R. and Herlocker, Jonathan L.}, biburl = {https://www.bibsonomy.org/bibtex/2340b69104710f8162ba10f9e4d54b2d9/lefteris8}, booktitle = {SIGIR '04: Proceedings of the 27th annual international conference on Research and development in information retrieval}, citeulike-article-id = {147672}, comment = {This paper pointed out with some shortcomings of traditional accuracy based performance metric commonly used in CF. Measures like MAE do not consider top items more important than low items in the recommendation list. This motivation was also noted in our ranking cf works. In the future, we should add this reference when making this arguments. Their experiments find that algorithms such as KNN predicts poorly for top items though overall accrucy measured by MAE still appear to be good. The proposed new evaluation procedure by resorting to IR metrics such as precision and recall and treat rated items as relevant and unrated items as irrelevant. A new algorithm is also proposed and shown to outperform baselines.}, doi = {10.1145/1008992.1009050}, interhash = {0996f4d7a8bf8642a9f3d671b2234c20}, intrahash = {340b69104710f8162ba10f9e4d54b2d9}, isbn = {1581138814}, keywords = {collaborative_filtering evaluation}, pages = {329--336}, posted-at = {2009-04-17 03:44:32}, priority = {2}, publisher = {ACM Press}, timestamp = {2009-06-22T17:28:39.000+0200}, title = {A collaborative filtering algorithm and evaluation metric that accurately model the user experience}, url = {http://dx.doi.org/10.1145/1008992.1009050}, year = 2004 }