融入专业度和用户相似性的跨域推荐算法
|
摘要
针对基于聚类型矩阵分解的跨域推荐算法过于泛化的问题,通过辅助领域中专业度高的相似用户在目标领域的评分来缓解跨域粗矩阵预测评分的泛化性,提出一种融入专业度和用户相似性的跨域推荐算法。在辅助领域计算冷启动用户与重叠用户的相似度,利用专业度高的相似用户在目标领域的评分预测其个性化评分,通过多域联合矩阵的分解聚类得到泛化的预测评分,两者进行加权综合,得到准确性较高的推荐项目。真实数据集上的实验结果表明,通过融入专业度和用户相似性,提出算法提高了冷启动用户的推荐准确性。
Aiming at the problem of over-generalization of cross-domain recommendation algorithm based on clustering matrix decomposition,the generalization of the cross-domain rough matrix prediction score was improved by the similar users with high professionalism in the target field,so that a cross-domain recommendation algorithm was proposed that combined professionalism and user similarity.The overlap user's similarity was calculated in the auxiliary field,high degree of similar users' personalized score with high professionalism in the target area was used to predict for target users,and clustering decomposition of the crossdomain matrix was used to obtain generalized prediction score,the weighted personalized score was then got,improving the accuracy of the recommendation.Results of experiments on real data sets show that the proposed algorithm improves the recommendation accuracy for cold start users by integrating professionalism and user similarity.
Aiming at the problem of over-generalization of cross-domain recommendation algorithm based on clustering matrix decomposition,the generalization of the cross-domain rough matrix prediction score was improved by the similar users with high professionalism in the target field,so that a cross-domain recommendation algorithm was proposed that combined professionalism and user similarity.The overlap user's similarity was calculated in the auxiliary field,high degree of similar users' personalized score with high professionalism in the target area was used to predict for target users,and clustering decomposition of the crossdomain matrix was used to obtain generalized prediction score,the weighted personalized score was then got,improving the accuracy of the recommendation.Results of experiments on real data sets show that the proposed algorithm improves the recommendation accuracy for cold start users by integrating professionalism and user similarity.
引文
[1]WANG Jun,LI Shijun,YANG Sha,et al.A new transfer learning model for cross-domain recommendations[J].Chinese Journal of Computers,2017,40(10):2367-2380(in Chinese).[王俊,李石君,杨莎,等.一种新的用于跨领域推荐的迁移学习模型[J].计算机学报,2017,40(10):2367-2380.]
[2]Tomeo P,Noia TD,Sciascio ED.Accuracy and diversity in cross-domain recommendations for cold-start users with positiveonly feedback[C]//ACM Conference on Recommender Systems.Boston,USA,2016:119-122.
[3]Tan Shulong,Bu Jiajun,Qin Xuzhen.Cross domain recommendation based on multi-type media fusion[J].Neurocomputing,2014,127(3):124-134.
[4]Chen W,Hsu W,Lee ML. Making recommendations from multiple domains[C]//ACM Sigkdd International Conference on Knowledge Discovery&Data Mining.Chicago, USA,2013:892-900.
[5]Fang YN,Guo YF,Ding XT,et al.An improved regularized singular value decomposition recommender algorithm based on tag transfer learning[J].Journal of Electronics&Information Technology,2013,35(12):3046-3050.
[6]Jiang M,Cui P,Chen X.Social recommendation with crossdomain transferable knowledge[J].IEEE Transactions on Knowledge&Data Engineering,2015,27(11):3084-3097.
[7]Nima Mirbakhsh,Charles X Ling.Improving Top-N recommendation for cold-start users via cross-domain information[J].ACM Transactions on Knowledge Discovery from Data,2015,9(4):19-38.
[8]YANG Changchun,SUN Jing.New social recommendation model of random walks based on users groups relation mining[J].Journal of Chinese Computer Systems,2012,33(3):565-570(in Chinese).[杨长春,孙婧.一种新的基于用户群体关系挖掘的随机漫游社会推荐模型[J].小型微型计算机系统,2012,33(3):565-570.]
[9]YU Chunhua,LIU Xuejun,LI Bin,et al.Mobile service recommendation based on context similarity and social network[J].Acta Electronica Sinica,2017,45(6):1530-1536(in Chinese).[俞春花,刘学军,李斌,等.基于上下文相似度和社会网络的移动服务推荐方法[J].电子学报,2017,45(6):1530-1536.]
[10]Ocepek U,Rugelj J,Bosnic Z.Improving matrix factorization recommendations for examples in cold start[J].Expert System with Applications,2015,42(19):6784-6794.
[11]Shi L,Zhao WX,Shen YD.Local representative-based matrix factorization for cold-start recommendation[J].ACM Transactions on Information Systems,2017,36(2):1-28.
[12]JING Minchang,TANG Diguan,YU Yinghui.A recommendation method based on expert trust in collaborative filtering[J].Library and Information Service,2012,56(11):105-108(in Chinese).[景民昌,唐弟官,于迎辉.基于专家信任优先的协同过滤推荐算法[J].图书情报工作,2012,56(11):105-108.]
[13]Hu Liang,Cao Jian, Xu Guandong,et al.Personalized recommendation via cross-domain triadic factorization[C]//Proceedings of the 22nd International Conference on World Wide Web,2013:595-606.
[2]Tomeo P,Noia TD,Sciascio ED.Accuracy and diversity in cross-domain recommendations for cold-start users with positiveonly feedback[C]//ACM Conference on Recommender Systems.Boston,USA,2016:119-122.
[3]Tan Shulong,Bu Jiajun,Qin Xuzhen.Cross domain recommendation based on multi-type media fusion[J].Neurocomputing,2014,127(3):124-134.
[4]Chen W,Hsu W,Lee ML. Making recommendations from multiple domains[C]//ACM Sigkdd International Conference on Knowledge Discovery&Data Mining.Chicago, USA,2013:892-900.
[5]Fang YN,Guo YF,Ding XT,et al.An improved regularized singular value decomposition recommender algorithm based on tag transfer learning[J].Journal of Electronics&Information Technology,2013,35(12):3046-3050.
[6]Jiang M,Cui P,Chen X.Social recommendation with crossdomain transferable knowledge[J].IEEE Transactions on Knowledge&Data Engineering,2015,27(11):3084-3097.
[7]Nima Mirbakhsh,Charles X Ling.Improving Top-N recommendation for cold-start users via cross-domain information[J].ACM Transactions on Knowledge Discovery from Data,2015,9(4):19-38.
[8]YANG Changchun,SUN Jing.New social recommendation model of random walks based on users groups relation mining[J].Journal of Chinese Computer Systems,2012,33(3):565-570(in Chinese).[杨长春,孙婧.一种新的基于用户群体关系挖掘的随机漫游社会推荐模型[J].小型微型计算机系统,2012,33(3):565-570.]
[9]YU Chunhua,LIU Xuejun,LI Bin,et al.Mobile service recommendation based on context similarity and social network[J].Acta Electronica Sinica,2017,45(6):1530-1536(in Chinese).[俞春花,刘学军,李斌,等.基于上下文相似度和社会网络的移动服务推荐方法[J].电子学报,2017,45(6):1530-1536.]
[10]Ocepek U,Rugelj J,Bosnic Z.Improving matrix factorization recommendations for examples in cold start[J].Expert System with Applications,2015,42(19):6784-6794.
[11]Shi L,Zhao WX,Shen YD.Local representative-based matrix factorization for cold-start recommendation[J].ACM Transactions on Information Systems,2017,36(2):1-28.
[12]JING Minchang,TANG Diguan,YU Yinghui.A recommendation method based on expert trust in collaborative filtering[J].Library and Information Service,2012,56(11):105-108(in Chinese).[景民昌,唐弟官,于迎辉.基于专家信任优先的协同过滤推荐算法[J].图书情报工作,2012,56(11):105-108.]
[13]Hu Liang,Cao Jian, Xu Guandong,et al.Personalized recommendation via cross-domain triadic factorization[C]//Proceedings of the 22nd International Conference on World Wide Web,2013:595-606.