Volume 95, Number 5, September 2011
|Number of page(s)||6|
|Section||Interdisciplinary Physics and Related Areas of Science and Technology|
|Published online||18 August 2011|
An item-oriented recommendation algorithm on cold-start problem
School of Information Engineering, Nanchang Hangkong University - Nanchang 330063, PRC
2 Web Sciences Center, University of Electronic Science and Technology of China - Chengdu 610054, PRC
3 Institute of Information Economy, Hangzhou Normal University - Hangzhou 310036, PRC
4 Department of Physics, University of Fribourg - Chemin du Musée, 1700 Fribourg, Switzerland
5 Department of Modern Physics, University of Science and Technology of China - Hefei 230026, PRC
Accepted: 13 July 2011
Based on a hybrid algorithm incorporating the heat conduction and probability spreading processes (Proc. Natl. Acad. Sci. U.S.A., 107 (2010) 4511), in this letter, we propose an improved method by introducing an item-oriented function, focusing on solving the dilemma of the recommendation accuracy between the cold and popular items. Differently from previous works, the present algorithm does not require any additional information (e.g., tags). Further experimental results obtained in three real datasets, RYM, Netflix and MovieLens, show that, compared with the original hybrid method, the proposed algorithm significantly enhances the recommendation accuracy of the cold items, while it keeps the recommendation accuracy of the overall and the popular items. This work might shed some light on both understanding and designing effective methods for long-tailed online applications of recommender systems.
PACS: 89.20.Ff – Computer science and technology / 89.75.Hc – Networks and genealogical trees / 89.65.-s – Social and economic systems
© EPLA, 2011
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.