一种基于信任机制的概率矩阵分解协同过滤推荐算法
|
摘要
传统的概率矩阵分解算法虽然较好地解决了推荐过程中的稀疏性和用户冷启动问题,但由于没有考虑到用户之间的信任信息,造成推荐精度不高.本文利用用户评分过程中潜在存在的信任关系,提出一种基于信任机制的概率矩阵分解协同过滤推荐算法TM-PMF(Probabilistic Matrix Factorization Algorithm of Collaborative Filtering Based on Trust Mechanism).首先根据用户间的信任关系来构建信任网络以获得信任评分矩阵.然后将信任评分矩阵与用户评分矩阵进行融合构建用户-信任评分矩阵,接着通过概率矩阵分解技术获得最优推荐列表.最终实验结果表明在不同稀疏数据集上,本文提出的TM-PMF算法较传统算法在精度方面有较大幅度地提高.
Although the Probabilistic Matrix Factorization algorithm is better to alleviate the data sparsity and cold start that every collaborative filtering algorithm or recommendation system confronts,due to ignore the trust information among users,the prediction accuracy is not high. In this paper,a probabilistic matrix factorization algorithm of collaborative filtering based on trust mechanism( TM-PMF) is proposed by employing both users' potential trust relationship and rating records. First,the trust network is constructed according to the trust relationship,to obtain the trust score matrix. Second,the trust score matrix is merged with the user-item rating matrix to construct the user-trust matrix. Finally,the optimal recommendation list is obtained by the Probabilistic Matrix Factorization technique. Experimental results showthat the algorithm proposed in this paper is more accurate than the traditional algorithm in different sparse data sets,especially for the Probabilistic Matrix Factorization model.
Although the Probabilistic Matrix Factorization algorithm is better to alleviate the data sparsity and cold start that every collaborative filtering algorithm or recommendation system confronts,due to ignore the trust information among users,the prediction accuracy is not high. In this paper,a probabilistic matrix factorization algorithm of collaborative filtering based on trust mechanism( TM-PMF) is proposed by employing both users' potential trust relationship and rating records. First,the trust network is constructed according to the trust relationship,to obtain the trust score matrix. Second,the trust score matrix is merged with the user-item rating matrix to construct the user-trust matrix. Finally,the optimal recommendation list is obtained by the Probabilistic Matrix Factorization technique. Experimental results showthat the algorithm proposed in this paper is more accurate than the traditional algorithm in different sparse data sets,especially for the Probabilistic Matrix Factorization model.
引文
[1] Chen Li,Chen Guan-liang,Wang Feng. Recommender systems based on user reviews:the state of the art[J]. User Modeling and User-Adapted Interaction,2015,25(2):99-154.
[2]Mirko Polato,Fabio Aiolli. Exploiting sparsity to build efficient kernel based collaborative filtering for top-N item recommendation[J]. Neurocomputing,2016,268(1):17-26.
[3]Uro2 Ocepek,Jo6e Rugelj,Zoran Bosnic'. Improving matrix factorization recommendations for examples in cold start[J]. Expert Systems with Applications,2015,42(19):6784-6794.
[4] Farhad Zafari,Irene Moser. Feature-aware factorised collaborative filtering[C]. In:Kang B.,Bai Q. ed. AI 2016:Advances in Artificial Intelligence,Proceedings of the Australasian Joint Conference on Artificial Intelligence,2016:561-569.
[5]Farhad Zafari,Irene Moser. Modelling socially-influenced conditional preferences over feature values in recommender systems based on factorised collaborative filtering[J]. Expert Systems with Applications,2017,87(1):98-117.
[6]Dawn Delay,Zhang Lin-lin,Laura D Hanish,et al. Peer influence on academic performance:a social network analysis of social-emotional intervention effects[J]. Prevention Science the Official Journal of the Society for Prevention Research,2016,17(8):1-11.
[7] Reza Zafarani,Mohammad Ali Abbasi,LIU Huan. Social media mining:an introduction[M]. NewYork:Cambridge University Press,2014.
[8] Salakhutdinov Ruslan,Mnih Andriy. Probabilistic matrix factorization[C]. In:John C Platt,D. Koller,Y. Singer,Sam T Roweis ed.Advances in Neural Information Processing Systems 20,Proceedings of the 20th International Conference on Neural Information Processing Systems,2007:1257-1264.
[9]MinGun Kim,Kyoungjae Kim. Recommender systems using SVD with social network information[J]. Journal of Intelligence and Information Systems,2016,22(4):1-18.
[10]Jiang Wen-jun,Wang Guo-jun,Md Zakirul Alam Bhuiyan,et al.Understanding graph-based trust evaluation in online social networks:methodologies and challenges[J]. Acm Computing Surveys,2016,49(1):1-35.
[11]Chang Wei-lun,Arleen N Diaz,Patrick C K Hung. Estimating trust value:a social network perspective[J]. Information Systems Frontiers,2015,17(6):1381-1400.
[12]Guo Gui-bing,Zhang Jie,Daniel Thalmann,et al. From ratings to trust:an empirical study of implicit trust in recommender systems[C]. In:Bracha Shapira ed. Recommender Systems:Theory and Applications Track,Proceedings of the 29th Annual ACMSymposium on Applied Computing,2014:248-253.
[13]Paolo Massa,Bobby Bhattacharjee. Using trust in recommender systems:an experimental analysis[C]. In:Jensen C. ed. Trust Management,Proceedings of Lecture Notes in Computer Science,vol2995,2004:221-235.
[14]Parham Moradi,Sajad Ahmadian. A reliability-based recommendation method to improve trust-aware recommender systems[J]. Expert Systems with Applications,2015,42(21):7386-7398.
[15]Ma Hao,Yang Hai-xuan,Lyu Michael R,et al. SoRec:social recommendation using probabilistic matrix factorization[C]. In:Rakesh Aqrawal ed. IR:Recommender Systems,Proceedings of the17th ACMConference on Information and Knowledge Management,2008:931-940.
[16]Mohsen Jamali,Martin Ester. A matrix factorization technique with trust propagation for recommendation in social networks[C]. In:Dietmar Jannnach ed. Recommending in Social Networks,Proceedings of the Fourth ACMConference on Recommender Systems,2010:135-142.
[17]Guo Gui-bing,Zhang Jie,Neil Yorke-Smith. A novel bayesian similarity measure for recommender systems[C]. In:Francesca Rossi ed. Web and Knowledge-based Information Systems,Proceedings of the 23th International Joint Conference on Artifical Intelligence,2013:2619-2625.
[2]Mirko Polato,Fabio Aiolli. Exploiting sparsity to build efficient kernel based collaborative filtering for top-N item recommendation[J]. Neurocomputing,2016,268(1):17-26.
[3]Uro2 Ocepek,Jo6e Rugelj,Zoran Bosnic'. Improving matrix factorization recommendations for examples in cold start[J]. Expert Systems with Applications,2015,42(19):6784-6794.
[4] Farhad Zafari,Irene Moser. Feature-aware factorised collaborative filtering[C]. In:Kang B.,Bai Q. ed. AI 2016:Advances in Artificial Intelligence,Proceedings of the Australasian Joint Conference on Artificial Intelligence,2016:561-569.
[5]Farhad Zafari,Irene Moser. Modelling socially-influenced conditional preferences over feature values in recommender systems based on factorised collaborative filtering[J]. Expert Systems with Applications,2017,87(1):98-117.
[6]Dawn Delay,Zhang Lin-lin,Laura D Hanish,et al. Peer influence on academic performance:a social network analysis of social-emotional intervention effects[J]. Prevention Science the Official Journal of the Society for Prevention Research,2016,17(8):1-11.
[7] Reza Zafarani,Mohammad Ali Abbasi,LIU Huan. Social media mining:an introduction[M]. NewYork:Cambridge University Press,2014.
[8] Salakhutdinov Ruslan,Mnih Andriy. Probabilistic matrix factorization[C]. In:John C Platt,D. Koller,Y. Singer,Sam T Roweis ed.Advances in Neural Information Processing Systems 20,Proceedings of the 20th International Conference on Neural Information Processing Systems,2007:1257-1264.
[9]MinGun Kim,Kyoungjae Kim. Recommender systems using SVD with social network information[J]. Journal of Intelligence and Information Systems,2016,22(4):1-18.
[10]Jiang Wen-jun,Wang Guo-jun,Md Zakirul Alam Bhuiyan,et al.Understanding graph-based trust evaluation in online social networks:methodologies and challenges[J]. Acm Computing Surveys,2016,49(1):1-35.
[11]Chang Wei-lun,Arleen N Diaz,Patrick C K Hung. Estimating trust value:a social network perspective[J]. Information Systems Frontiers,2015,17(6):1381-1400.
[12]Guo Gui-bing,Zhang Jie,Daniel Thalmann,et al. From ratings to trust:an empirical study of implicit trust in recommender systems[C]. In:Bracha Shapira ed. Recommender Systems:Theory and Applications Track,Proceedings of the 29th Annual ACMSymposium on Applied Computing,2014:248-253.
[13]Paolo Massa,Bobby Bhattacharjee. Using trust in recommender systems:an experimental analysis[C]. In:Jensen C. ed. Trust Management,Proceedings of Lecture Notes in Computer Science,vol2995,2004:221-235.
[14]Parham Moradi,Sajad Ahmadian. A reliability-based recommendation method to improve trust-aware recommender systems[J]. Expert Systems with Applications,2015,42(21):7386-7398.
[15]Ma Hao,Yang Hai-xuan,Lyu Michael R,et al. SoRec:social recommendation using probabilistic matrix factorization[C]. In:Rakesh Aqrawal ed. IR:Recommender Systems,Proceedings of the17th ACMConference on Information and Knowledge Management,2008:931-940.
[16]Mohsen Jamali,Martin Ester. A matrix factorization technique with trust propagation for recommendation in social networks[C]. In:Dietmar Jannnach ed. Recommending in Social Networks,Proceedings of the Fourth ACMConference on Recommender Systems,2010:135-142.
[17]Guo Gui-bing,Zhang Jie,Neil Yorke-Smith. A novel bayesian similarity measure for recommender systems[C]. In:Francesca Rossi ed. Web and Knowledge-based Information Systems,Proceedings of the 23th International Joint Conference on Artifical Intelligence,2013:2619-2625.