基于矩阵分解的社会化推荐模型

严素蓉; 冯小青; 廖一星

doi:10.16511/j.cnki.qhdxxb.2016.21.045

清华大学学报（自然科学版） >

2016 , Vol. 56 >Issue 7: 793 - 800

DOI: https://doi.org/10.16511/j.cnki.qhdxxb.2016.21.045

计算机科学与技术

基于矩阵分解的社会化推荐模型

严素蓉 ,
冯小青 ,
廖一星

展开

1. 浙江财经大学信息学院, 杭州 310018;
2. 加州大学尔湾分校电子工程与计算机科学系, 加利福尼亚 92697-2625

收稿日期: 2015-09-11

网络出版日期: 2016-07-15

基金资助

国家自然科学青年基金项目（61502414，61202197）；浙江省自然科学青年基金项目（LQ14F010006）

收起

Matrix factorization based social recommender model

YAN Surong ,
FENG Xiaoqing ,
LIAO Yixing

Expand

1. College of Information, Zhejiang University of Finance and Economics, Hangzhou 310018, China;
2. Department of Electrical Engineering & Computer Science, University of California, Irvine, California 92697-2625, USA

Received date: 2015-09-11

Online published: 2016-07-15

Fold

摘要

该文提出一种经由定制关系网络改进的基于矩阵分解的社会化推荐模型来缓解数据稀疏性和冷启动问题，并进一步改善大数据集导致的可扩展性问题。在该模型中，关系网络的社交影响力被建模为矩阵分解模型的用户-物品（user-item）评分倾向，而同质性则被建模为动态正则项。为了获得更好的预测精度和可扩展性，设计了一个关系网络boosting-shrinking算法，在该算法中，基于用户在数据集中的数据密度，自适应地裁减每个用户的关系网络为其定制个性化的关系网络。在稀疏水平不同的不平衡数据集上的实验表明：相比其他的基于矩阵分解的社会化推荐模型，该模型可以显著提高稀疏数据集的预测精度，有效地缓解冷启动问题，并获得较好的可扩展性。

关键词： 大数据; 社交网络; 矩阵分解; 稀疏性; 可扩展性

本文引用格式

严素蓉 , 冯小青 , 廖一星 . 基于矩阵分解的社会化推荐模型[J]. 清华大学学报（自然科学版）, 2016 , 56(7) : 793 -800 . DOI: 10.16511/j.cnki.qhdxxb.2016.21.045

Abstract

This study describes an improved matrix factorization based social recommender model that uses tailored relationship networks of users as a solution for the sparsity, cold-start and scalability problems in big datasets. The social influence of the relationship networks is targeted as an extra user-item specific bias for the matrix factorization with the uniformity of relationship networks modeled as dynamic social regularization terms in the matrix factorization. A boosting-shrinking algorithm is used for the relationship networks for better prediction accuracy and scalability where the relationships of each user are tailored to generate personalized relationship networks according to the user-specific data density of the user-item rating matrix and the correlation matrix. Tests on unbalanced datasets with different sparsity levels show that this model significantly improves the prediction accuracy for sparse datasets, effectively addresses the cold-start problem, and has better scalability compared to other state-of-the-art matrix factorization based social recommendation models.

Key words： big data; social networks; matrix factorization; sparsity; scalability

参考文献

[1] Koren Y. Factorization meets the neighborhood:A multifaceted collaborative filtering model[C]//Proc ACM SIGKDD'08. Las Vegas, NE, USA:ACM, 2008:426-434.
[2] Su X, Khoshgoftaar T. A survey of collaborative filtering techniques[J]. Advances in Artificial Intelligence, 2009(2009), 421425.
[3] Tavakolifard M, Almeroth K C. Social computing:an intersection of recommender systems, trust/reputation systems, and social networks[J]. Network, IEEE, 2012, 26(4):53-58.
[4] Xin J C, Wang Z Q, Qu L X, et al. Elastic extreme learning machine for big data classification[J]. Neuro computing, 2015(149):464-471.
[5] Jamiy E L, Daif A, Azouazi M, et al. The potential and challenges of Big data-Recommendation systems next level application[J]. International Journal of Computer Science Issues, 2014, 11(5):21-26.
[6] Pagare R, Patil S A. Study of collaborative filtering recommendation algorithm-scalability issue[J]. International Journal of Computer Applications, 2013, 67(25):10-15
[7] McPherson M, Smith-Lovin L, Cook J. Birds of a feather:Homophily in social networks[J]. Annual review of sociology, 2001, 27:415-444.
[8] Marsden P, Friedkin N. Network studies of social influence[J]. Sociological Methods and Research, 1993, 22(1):127-151.
[9] Liu LY, Medob M, Yeung CH, et al. Recommender systems[J]. Physics Reports, 2012, 59(1):1-49.
[10] Ma H, Yang H, Lyu M R, et al. SoRec:Social recommendation using probabilistic matrix factorization[C]//Proc ACM IKM'08. Napa Valley, CA, USA:ACM, 2008:931-940.
[11] Ma H, King I, Lyu M R. Learning to recommend with social trust ensemble[C]//Proc ACM SIGIR'09. Boston, MA, USA:2009:203-210.
[12] Ma H, Zhou D, Liu C, et al. Recommender systems withsocial regularization[C]//Proc ACM WSMD'11. Hong Kong, China:ACM, 2011:287-296.
[13] Jamali M, Ester M. Trustwalker:a random walk model for combining trust-based and item-based recommendation[C]//Proc ACM SIGKDD'09. Paris, France:ACM, 2009:397-406.
[14] Jamali M, Ester M. A matrix factorization technique with trust propagation for recommendation in social networks[C]//Proc ACM RecSys'10. Barcelona, Spain:ACM, 2010:135-142.
[15] Symeonidis P, Tiakas E, Manolopoulos Y. Product recommendation and rating prediction based on multi-modal social networks[C]//Proc ACM RecSys'11. Chicago, IL, USA:ACM, 2011:61-68.
[16] Yuan Q, Chen L, Zhao S. Factorization vs. regularization:fusing heterogeneous social relationships in top-n recommendation[C]//Proc ACM RecSys'11. Chicago, IL, USA:ACM, 2011:245-252.
[17] Yang X, Steck H, Liu Y. Circle-based recommendation in online social networks[C]//Proc ACM SIGKDD'12. Beijing, China:ACM, 2012:1267-1275.
[18] Noel J, Sanner S, Tran K N, et al. Objective functions for social collaborative filtering[C]//Proc WWW'12. Lyon, France, 2012:859-868.
[19] Yan S R, Zheng X L, Chen D R, et al. Exploiting two-faceted web of trust for enhanced-quality recommendations[J]. Expert Systems with Applications, 2013, 40(17):7080-7095.
[20] Chen C, Zheng X, Wang Y, et al. Context-aware collaborative topic regression with social matrix factorization for recommender systems[C]//Proc AAAI'14. Québec City, QU, Canada, 2014:9-15.
[21] Tang J, Hu X, Liu H. Social recommendation:A review[J]. Social network analysis and mining, 2013, 3(4):1113-1133.
[22] Colombo-Mendoza L O, Valencia-García R, Rodríguez-González A, et al. RecomMetz:A context-aware knowledge-based mobile recommender system for movie show times[J]. Expert Systems with Applications, 2015, 42(3):1202-1222.
[23] Liu X, Aberer K. SoCo:A social network aided context-aware recommender system[C]//Proc www'13. Rio de Janeiro, Brazil, 2013:781-802.
[24] Koren Y, Bell R, Volinsky C. Matrix factorization techniques for recommender systems[J]. Computer, 2009, 42(8):30-37.
[25] Salakhutdinov R, Mnih A. Probabilistic matrix factorization[J]. Advances in neural information processing systems, 2008, 20(1):1257-1264.
[26] Menon A K, Elkan C. A log-linear model with latent features for dyadic prediction[C]//Proc ICDM'10. Sydney, Australia:IEEE, 2010:364-373.
[27] Yan S R, Zheng X L, Chen D R, et al. User-centric trust and reputation model for personal and trusted service selection[J]. International Journal of Intelligent Systems, 2011, 26(8):687-717.
[28] Yan S R, Zheng X L, Wang Y, et al. Graph-based comprehensive reputation model:Exploiting the social context of opinions to enhance trust in social commerce[J]. Information Sciences, 2015, 318:51-72.
[29] Gantner Z, Rendle S, Freudenthaler C, et al. MyMediaLite:A Free Recommender System Library[C]//Proc ACM RecSys'11. Chicago, IL, USA:ACM, 2011:305-308.
[30] Wang J, Zhang Y. Utilizing marginal net utility for recommendation in e-commerce[C]//Proc ACM SIGIR'11. Beijing, China:ACM, 2011:1003-1012.
[31] Wang J, Zhang Y. Opportunity model for e-commerce recommendation:Right product; right time[C]//Proc ACM SIGIR' 13. Dublin, Ireland:ACM, 2013:303-312.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献

访问统计