融合项目属性偏好的矩阵分解推荐模型

doi:10.19665/j.issn1001-2400.2022.03.017

摘要/Abstract

摘要：

为了解决传统协同过滤算法针对数据稀疏,特别是冷启动等一系列问题时,无法准确地计算出用户与用户、物品与物品之间的相似度,进而无法精准地为用户推荐相应物品的难题,结合基于近邻的协同过滤算法及基于模型协同过滤算法的优势,提出了一种基于矩阵分解的推荐模型。该模型使用基于模型的协同过滤,以矩阵分解为基础,同时融入其他辅助信息,以期优化矩阵分解的效果,从而进行更精准的评分预测。基于传统矩阵分解算法,在已有的推荐模型中,首先基于用户属性与项目属性信息进行相似度计算,构建评分矩阵,进行用户的初始评分预测;然后融合用户对项目属性的喜好构建用户兴趣矩阵,同时以用户属性信息、项目属性信息作为辅助,融入到新的矩阵分解模型中,进行冷启动用户的评分预测。与传统的个性化推荐模型相比,新模型有着更好的推荐准确性。通过仿真实验,也证实了这个推荐模型对于冷启动问题有一定程度的缓解,准确性也有所提升。同时,在模型可扩展性等方面,也取得了较好的效果。

关键词: 用户冷启动, 矩阵分解, 用户属性, 项目属性, 用户偏好

Abstract:

In order to solve a series of problems such as sparse data,especially cold start,the traditional collaborative filtering algorithm can not accurately calculate the similarity between users and articles,so that it cannot accurately recommend corresponding articles for users.Combined with the advantages of the nearest neighbor based collaborative filtering algorithm and model-based collaborative filtering algorithm,this paper proposes a recommendation model based on matrix decomposition.The model uses model-based collaborative filtering,which is based on matrix decomposition and integrates other auxiliary information,in order to optimize the effect of matrix decomposition and make more accurate score prediction.Based on the traditional matrix decomposition algorithm,in the existing recommendation model,first,the similarity is calculated based on the user attribute and project attribute information,and the scoring matrix is constructed to predict the user's initial score.Then,the user interest matrix is constructed by integrating the user’s preferences for project attributes.At the same time,the user attribute information and project attribute information are integrated into the new matrix decomposition model to predict the score of cold start users.Compared with the traditional personalized recommendation model,the new model has a better recommendation accuracy.Through simulation experiments,it is also confirmed that the recommended model mentioned in this paper can alleviate the cold start problem to a certain extent and improve the accuracy.At the same time,good results have been achieved in the scalability of the model.

Key words: user cold start, matrix decomposition, user attributes, project attributes, user preference

中图分类号:

TP391

韩立锋,陈莉,史晓龙. 融合项目属性偏好的矩阵分解推荐模型[J]. 西安电子科技大学学报, 2022, 49(3): 147-159.

HAN Lifeng,CHEN Li,SHI Xiaolong. Matrix decomposition recommendation model incorporating item attribute preference[J]. Journal of Xidian University, 2022, 49(3): 147-159.

图/表 12

表1

图1

图2

表2

表3

图3

表4

图4

图5

图6

图7

图8

参考文献 28

[1]	VERBERT K, MANOUSELIS N, OCHOA X, et al. Context-Aware Recommender Systems for Learning:A Survey and Future Challenges[J]. IEEE Transactions on Learning Technologies, 2012, 5(4):318-335. doi: 10.1109/TLT.2012.11
[2]	BELLOGíN A, SAID A. Information Retrieval and Recommender Systems (Data Science in Practice)[M]. Heidelberg:Springer, 2019.
[3]	LEAMING P M, MACKEY L. Collaborative Filtering[J]. Computer ence, 2017, 57(4):189.
[4]	LINDEN G, SMITH B, YORK J, et al. Amazon.com Recommendations:Item-to-Item Collaborative Filtering[J]. IEEE Internet Computing, 2003, 7(1):76-80. doi: 10.1109/MIC.2003.1167344
[5]	LEE S K, CHO Y H, KIM S H. Collaborative Filtering with Ordinal Scale-Based Implicit Ratings for Mobile Music Recommendations[J]. Information Sciences, 2010, 180(11):2142-2155. doi: 10.1016/j.ins.2010.02.004
[6]	SU J H, CHIU T W. An Item-Based Music Recommender System Using Music Content Similarity[C]// Asian Conference on Intelligent Information and Database Systems.Heidelberg:Springer, 2016:179-190.
[7]	LI W, ZHOU X, SHIMIZU S, et al. Personalization Recommendation Algorithm Based on Trust Correlation Degree and Matrix Factorization[J]. IEEE Access, 2019, 7:45451-45459. doi: 10.1109/ACCESS.2018.2885084
[8]	BOBADILLA J, SERRADILLA F, BERNAL J. A New Collaborative Filtering Metric that Improves The Behavior of Recommender Systems[J]. Knowledge-Based Systems, 2010, 23(6):520-528. doi: 10.1016/j.knosys.2010.03.009
[9]	LIU H, HU Z, MIAN A, et al. A New User Similarity Model to Improve the Accuracy of Collaborative Filtering[J]. Knowledge-Based Systems, 2014, 56:156-166. doi: 10.1016/j.knosys.2013.11.006
[10]	LI B, ZHU X, LI R, et al. Rating Knowledge Sharing in Cross-Domain Collaborative Filtering[J]. IEEE Transactions on Cybernetics, 2017, 45(5):1068-1082. doi: 10.1109/TCYB.2014.2343982
[11]	KOREN Y, BELL R, VOLINSKY C. Matrix Factorization Techniques for Recommender Systems[J]. Computer, 2009, 42(8):30-37.
[12]	MANZATO M. Discovering Latent Factors from Movies Genres for Enhanced Recommendation[C]// Proceedings of the 6th ACM Conference on Recom-mender Systems. New York: ACM, 2012:249-252.
[13]	QIN J, CAO L, PENG H. Collaborative Filtering Recommendation Algorithm Based on Weighted Item Category[C]// 2016 Chinese Control and Decision Conference (CCDC).Piscataway:IEEE, 2016:2782-2786.
[14]	CHEN C, ZHANG M, LIU Y, et al. Neural Attentional Rating Regression with Review-Level Explanations[C]// Proceedings of the 2018 World Wide Web Conference. New York: ACM, 2018:1583-1592.
[15]	XUE F HE X, WANG X, et al. Deep Item-Based Collaborative Filtering for Top-N Recommendation[J/OL]. [2020-12-20]. DOI: 10.48550/arXiv.1811.04392. doi: 10.48550/arXiv.1811.04392
[16]	HE X, HE Z, DU X, et al. Adversarial Personalized Ranking for Recommendation[C/OL]. [2020-12-22]. https://www.xueshufan.com/publication/3099988866.
[17]	WANG X, HE X, WANG M, et al. Neural Graph Collaborative Filtering[C/OL]. [2020-12-21]. Staff.ustc.edu.cn/-hexn/papers/sigir19-NGCF.pdf.
[18]	GOU J, LAN D, ZHANG Y, et al. A New Distance-Weighted k-Nearest Neighbor Classifier[J]. Journal of Information & Computational Science, 2012, 9(6):1429-1436.
[19]	YU Z, CHEN H, LIU J, et al. Hybrid k-Nearest Neighbor Classifier[J]. IEEE Transactions on Cybemetics, 2015, 46(6):1263-1275.
[20]	GUO X, YIN S C, ZHANGY W, et al. Cold Start Recommendation Based on Attribute-Fused Singular Value Decomposition[J]. IEEE Access, 2019, 7:11349-11359. doi: 10.1109/ACCESS.2019.2891544
[21]	JIANG B, LUO Y. Matrix Factorization Based Instrumental Variable Approach for Simultaneous Identification of Bi-Directional Path Models[J]. ISA Transactions, 2018, 79:73-82. doi: 10.1016/j.isatra.2018.04.018
[22]	CHEN S, PENG Y. Matrix Factorization for Recommendation with Explicit and Implicit Feedback[J]. Knowledge Based Systems, 2018, 158:109-117. doi: 10.1016/j.knosys.2018.05.040
[23]	CHEN C, ZHANG M, ZHANG Y, et al. Efficient Neural Matrix Factorization without Sampling for Recommendation[J]. ACM Transactions on Information Systems, 2020, 38(2):1-28.
[24]	PARVIN H, MORADIA P, ESMAEILI S, et al. A Scalable and Robust Trust-Based Nonnegative Matrix Factorization Recommender Using The Alternating Direction Method[J]. Knowledge-Based Systems, 2019, 166:92-107. doi: 10.1016/j.knosys.2018.12.016
[25]	DENG S, HUANG L, XU G, et al. On Deep Learning for Trust-Aware Recommendations in Social Networks[J]. IEEE Transactions on Neural Networks & Learning Systems, 2017, 28(5):1164-1177.
[26]	KOREN Y. Factorization Meets the Neighborhood:A Multifaceted Collaborative Filtering Model[C]// Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York: ACM, 2008:426-434.
[27]	SALAKHUTDINOV R, MNIH A. Probabilistic Matrix Factorization[C]// Proceedings of the 20th International Conference on Neural Information Processing Systems. New York: ACM, 2007:1257-1264.
[28]	NGUYEN J, ZHU M. Content-Boosted Matrix Factorization Techniques for Recommender Systems[J]. Statistical Analysis and Data Mining:The ASA Data Science Journal, 2013, 6(4):286-301. doi: 10.1002/sam.11184

用户	项目1	…	项目j	…	项目n
用户1	R_1,1	…	R_1,_j	…	R_1,_n
?	?		?		?
用户i	R_i_,1	…	R_i_,_j	…	R_i_,_n
?	?		?		?
用户m	R_m_,1	…	R_m_,_j	…	R_m_,_n

用户ID	年龄	性别	职业
1	2	1	2
2	6	0	7
3	2	1	4
?	?	?	?
943	2	1	5

电影名	类型	年代	地区	评分
电影A	浪漫	80年代	美国	3
电影B	儿童、动画	90年代	美国	5
电影C	动画	80年代	英国	4
电影D	动作、悬疑	90年代	日本	4

数据集名称	用户数	电影数	评分数
MovieLens ML-100K	943	1 682	100 004
MovieLens ML-1M	6 040	3 883	1 000 000