TY - JOUR
T1 - Adaptive ensemble with trust networks and collaborative recommendations
AU - Zou, Haitao
AU - Gong, Zhiguo
AU - Zhang, Nan
AU - Li, Qing
AU - Rao, Yanghui
N1 - Funding Information:
Haitao Zou and Zhiguo Gong were supported in part by Fund of Science and Technology Development of Macau Government under FDCT/106/2012/A3 and FDCT/116/2013/A3 and in part by University Macau Research Committee under MYRG188-FST11-GZG and MYRG105-FST13-GZG. Nan Zhang was supported in part by the National Science Foundation under Grants 0852674, 0915834, 1117297, and 1343976. Any opinions, findings, conclusions, and/or recommendations expressed in this material, either expressed or implied, are those of the authors and do not necessarily reflect the views of the sponsors listed above.
Publisher Copyright:
© 2014, Springer-Verlag London.
PY - 2015/9/17
Y1 - 2015/9/17
N2 - Several existing recommender algorithms combine collaborative filtering and social/trust networks together in order to overcome the problems caused by data scarcity and to produce more effective recommendations for users. In general, those methods fuse a user’s own taste and his trusted friends/users’ tastes using an ensemble model where a parameter is used to balance these two components. However, this parameter is often set as a constant and with no regard to users’ individual characteristics. Aiming at introducing personalization to solve the above problem, we propose a local topology-based ensemble model to adaptively combine a user’s own taste and his trusted friends/users’ tastes. We take users’ clustering coefficients in the social/trust networks as the indicator to measure the consistence of their selecting trusted friends/users and leverage this local topology-based parameter in the ensemble model. To predict the likelihood of users’ purchasing actions on items, we also combine item ratings and sentiment values which are reflected in the review contents as the input to the adaptive ensemble model. We conduct comprehensive experiments which demonstrate the superiority of our adaptive algorithms over the existing ones.
AB - Several existing recommender algorithms combine collaborative filtering and social/trust networks together in order to overcome the problems caused by data scarcity and to produce more effective recommendations for users. In general, those methods fuse a user’s own taste and his trusted friends/users’ tastes using an ensemble model where a parameter is used to balance these two components. However, this parameter is often set as a constant and with no regard to users’ individual characteristics. Aiming at introducing personalization to solve the above problem, we propose a local topology-based ensemble model to adaptively combine a user’s own taste and his trusted friends/users’ tastes. We take users’ clustering coefficients in the social/trust networks as the indicator to measure the consistence of their selecting trusted friends/users and leverage this local topology-based parameter in the ensemble model. To predict the likelihood of users’ purchasing actions on items, we also combine item ratings and sentiment values which are reflected in the review contents as the input to the adaptive ensemble model. We conduct comprehensive experiments which demonstrate the superiority of our adaptive algorithms over the existing ones.
UR - http://www.scopus.com/inward/record.url?scp=85027947809&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85027947809&partnerID=8YFLogxK
U2 - 10.1007/s10115-014-0782-7
DO - 10.1007/s10115-014-0782-7
M3 - Article
AN - SCOPUS:85027947809
SN - 0219-1377
VL - 44
SP - 663
EP - 688
JO - Knowledge and Information Systems
JF - Knowledge and Information Systems
IS - 3
ER -