基于内容和最近邻算法的多臂老虎机推荐算法
|
摘要
为有效解决推荐系统的冷启动问题和动态数据建模问题,基于多臂老虎机算法与协同过滤算法,利用用户信息反馈在线及时更新推荐模型;将推荐系统的冷启动问题转化成探索和利用(Explore&Exploit,简称E&E)问题,利用多臂老虎机算法,在引入用户特征为内容的基础上,进一步考虑用户之间的协同作用,提出基于内容和最近邻算法的多臂老虎机推荐算法;采用Movielens和Jester的真实数据集进行对比实验,实验结果表明:k NNUCB算法更优且更具实用性,尤其在解决冷启动问题上效果显著.
In order to solve the problems of cold start and dynamic data modeling,based on the combination of the multi-armed bandit algorithm and the collaborative filtering algorithm,the users' feedbacks can be used to timely update the model online. The cold start problem can be easily converted into Explore & Exploit( E&E) problems.Considering the synergy between users,a multi-armed bandit recommendation system is proposed based on contextual content and kNN( k-Nearest Neighbors) algorithm. The experiments on real datasets from Movielens and Jester are conducted. The experimental results show that the contextual multi-armed bandit recommendation system based on kNN performs better compared with other baseline approaches. And the proposed algorithm is particularly effective in solving the cold start problem.
In order to solve the problems of cold start and dynamic data modeling,based on the combination of the multi-armed bandit algorithm and the collaborative filtering algorithm,the users' feedbacks can be used to timely update the model online. The cold start problem can be easily converted into Explore & Exploit( E&E) problems.Considering the synergy between users,a multi-armed bandit recommendation system is proposed based on contextual content and kNN( k-Nearest Neighbors) algorithm. The experiments on real datasets from Movielens and Jester are conducted. The experimental results show that the contextual multi-armed bandit recommendation system based on kNN performs better compared with other baseline approaches. And the proposed algorithm is particularly effective in solving the cold start problem.
引文
[1]于洪,李俊华.一种解决新项目冷启动问题的推荐算法[J].软件学报,2015,26(6):1395-1408.YU H,LI J H. Algorithm to solve the cold-start problem in new item recommendations[J]. Journal of Software,2015,26(6):1395-1408.
[2]陈洁敏,汤庸,李建国,等.个性化推荐算法研究[J].华南师范大学学报(自然科学版),2014,46(5):8-15.CHEN J M,TANG Y,LI J G,et al. Survey of personalized recommendation algorithms[J]. Journal of South China Normal University(Natural Science Edition),2014,46(5):8-15.
[3] LINDEN G,SMITH B,YORK J. Amazon.com recommendations:item-to-item collaborative filtering[J]. IEEE Internet Computing,2003,7(1):76-80.
[4] KOREN Y,BELL R,VOLINSKY C. Matrix factorization techniques for recommender systems[J]. Computer,2009,42(8):30-37.
[5] MOHRI M. Multi-armed bandit algorithms and empirical evaluation[C]∥Proceedings of the 16th European Conference on Machine Learning. Berlin:Springer,2005:437-448.
[6] GITTINS J,GLAZEBROOK K,WEBER R,et al. Multiarmed bandit allocation indices[J]. Journal of the Operational Research Society,1989,40(12):1158-1159.
[7] AGARWAL D,CHEN B C,ELANGO P. Explore/Exploit schemes for web content optimization[C]∥Proceedings of the Ninth IEEE International Conference on Data Mining. Washington,DC:IEEE,2009:1-10.
[8] AUER P,CESA-BIANCHI N,FISCHER P. Finite-time analysis of the multiarmed bandit problem[J]. Machine Learning,2002,47(2/3):235-256.
[9] HAI T N,MARY J,PREUX P. Cold-start problems in recommendation systems via contextual-bandit algorithms[J/OL].(2014-05-29)[2018-01-10]. http:∥arxiv.org/abs/1405.7544.
[10] CESA-BIANCHI N. Multi-armed bandit problem[M].New York:Springer,2016.
[11] AUER P,CESA-BIANCHI N,FREUND Y,et al. The nonstochastic multi-armed bandit problem[J]. Siam Journal on Computing,2012,32(1):48-77.
[12] CHAPELLE O,LI L. An empirical evaluation of thompson sampling[C]∥Proceedings of the 24th Internation Conference on Neural Information Processing Systems. New York:Curran Associates,2012:2249-2257.
[13] SCHULZ E,KONSTANTINIDIS E,SPEEKENBRINK M.Putting bandits into context:how function learning supports decision making[J]. Journal of Experimental Psychology Learning Memory&Cognition,2018,44(6):927-943.
[14] LI L H,CHU W,LANGFORD J,et al. A contextual-bandit approach to personalized news article recommendation[C]∥Proceedings of the International Conference on World Wide Web. New York:ACM,2010:661-670.
[15] GENTILE C,LI S,ZAPPELLA G. Online clustering of bandits[C]∥Proceedings of the International Conference on Machine Learning. Beijing:ICML,2014:757-765.
[16] LI S,KARATZOGLOU A,GENTILE C. Collaborative filtering bandits[C]∥Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York:ACM,2016:539-548.
[17] GENTILE C,LI S,KAR P,et al. On context-dependent clustering of bandits[C]∥Proceedings of the International Conference on Machine Learning. Sydney:ICML,2017:1253-1262.
[2]陈洁敏,汤庸,李建国,等.个性化推荐算法研究[J].华南师范大学学报(自然科学版),2014,46(5):8-15.CHEN J M,TANG Y,LI J G,et al. Survey of personalized recommendation algorithms[J]. Journal of South China Normal University(Natural Science Edition),2014,46(5):8-15.
[3] LINDEN G,SMITH B,YORK J. Amazon.com recommendations:item-to-item collaborative filtering[J]. IEEE Internet Computing,2003,7(1):76-80.
[4] KOREN Y,BELL R,VOLINSKY C. Matrix factorization techniques for recommender systems[J]. Computer,2009,42(8):30-37.
[5] MOHRI M. Multi-armed bandit algorithms and empirical evaluation[C]∥Proceedings of the 16th European Conference on Machine Learning. Berlin:Springer,2005:437-448.
[6] GITTINS J,GLAZEBROOK K,WEBER R,et al. Multiarmed bandit allocation indices[J]. Journal of the Operational Research Society,1989,40(12):1158-1159.
[7] AGARWAL D,CHEN B C,ELANGO P. Explore/Exploit schemes for web content optimization[C]∥Proceedings of the Ninth IEEE International Conference on Data Mining. Washington,DC:IEEE,2009:1-10.
[8] AUER P,CESA-BIANCHI N,FISCHER P. Finite-time analysis of the multiarmed bandit problem[J]. Machine Learning,2002,47(2/3):235-256.
[9] HAI T N,MARY J,PREUX P. Cold-start problems in recommendation systems via contextual-bandit algorithms[J/OL].(2014-05-29)[2018-01-10]. http:∥arxiv.org/abs/1405.7544.
[10] CESA-BIANCHI N. Multi-armed bandit problem[M].New York:Springer,2016.
[11] AUER P,CESA-BIANCHI N,FREUND Y,et al. The nonstochastic multi-armed bandit problem[J]. Siam Journal on Computing,2012,32(1):48-77.
[12] CHAPELLE O,LI L. An empirical evaluation of thompson sampling[C]∥Proceedings of the 24th Internation Conference on Neural Information Processing Systems. New York:Curran Associates,2012:2249-2257.
[13] SCHULZ E,KONSTANTINIDIS E,SPEEKENBRINK M.Putting bandits into context:how function learning supports decision making[J]. Journal of Experimental Psychology Learning Memory&Cognition,2018,44(6):927-943.
[14] LI L H,CHU W,LANGFORD J,et al. A contextual-bandit approach to personalized news article recommendation[C]∥Proceedings of the International Conference on World Wide Web. New York:ACM,2010:661-670.
[15] GENTILE C,LI S,ZAPPELLA G. Online clustering of bandits[C]∥Proceedings of the International Conference on Machine Learning. Beijing:ICML,2014:757-765.
[16] LI S,KARATZOGLOU A,GENTILE C. Collaborative filtering bandits[C]∥Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval. New York:ACM,2016:539-548.
[17] GENTILE C,LI S,KAR P,et al. On context-dependent clustering of bandits[C]∥Proceedings of the International Conference on Machine Learning. Sydney:ICML,2017:1253-1262.