结合物品流行度的列表级矩阵因子分解算法

doi:10.11772/j.issn.1001-9081.2017123066

计算机应用 ›› 2018, Vol. 38 ›› Issue (7): 1877-1881.DOI: 10.11772/j.issn.1001-9081.2017123066

结合物品流行度的列表级矩阵因子分解算法

周瑞环, 赵宏宇

西南交通大学信息科学与技术学院, 成都 611756

收稿日期:2018-01-02 修回日期:2018-03-01 出版日期:2018-07-10 发布日期:2018-07-12
通讯作者: 周瑞环
作者简介:周瑞环(1993-),男,浙江温州人,硕士研究生,主要研究方向:个性化推荐算法、机器学习;赵宏宇(1971-),男,重庆人,副教授,博士,主要研究方向:模式识别、人工智能、信息理论与编码。

List-wise matrix factorization algorithm with combination of item popularity

ZHOU Ruihuan, ZHAO Hongyu

School of Information Science and Technology, Southwest Jiaotong University, Chengdu Sichuan 611756, China

Received:2018-01-02 Revised:2018-03-01 Online:2018-07-10 Published:2018-07-12

摘要/Abstract

摘要： 针对变形的奇异值分解（SVD++）算法的评分规则在模型训练和预测两个阶段的不一致问题和列表级矩阵因子分解（ListRank-MF）算法的Top-1排序概率在大量物品评分一样时排序概率一样的问题，提出一种结合物品流行度的列表级矩阵因子分解算法。首先，在评分规则中使用到的用户有过行为的物品集合中去除当前待评分物品；接着结合物品流行度改进Top-1排序概率；然后使用随机梯度下降算法求解目标函数并进行Top-N推荐。基于修正的SVD++评分规则，在MovieLens和Netflix数据集上比较了所提算法与目标函数为点级和列表级的SVD++算法。所提算法与列表级的SVD++算法相比，Top-N推荐准确率指标归一化折损累积增益（NDCG）值在MovieLens数据集上提高了5%~8%，在Netflix数据集上提高了1%左右。实验结果表明，所提算法能够有效提高Top-N推荐准确率。

关键词: 矩阵因子分解, Top-N推荐, 变形的奇异值分解(SVD++)算法, 物品流行度, 随机梯度下降

Abstract: For the difference of transmutative Singular Value Decomposition (SVD++) algorithm's rating rule in two stages of model training and prediction, and the same probability of List-wise Matrix Factorization (ListRank-MF) algorithm's Top-1 ranking probability caused by a large number of same rating items, an algorithm of list-wise matrix factorization combining with item popularity was proposed. Firstly, the current item to be rated was removed from the set of items that the user had used in the rating rule. Secondly, the item popularity was used to improve the Top-1 ranking probability. Then the stochastic gradient descent algorithm was used to solve the objective function and make Top-N recommendation. Based on the modified SVD++ rating rule, the proposed algorithm and the SVD++ algorithms whose objective functions are point-wise and list-wise were compared on MovieLens and Netflix datasets. Compared with the list-wise SVD++ algorithm, the value of Normalized Discounted Cumulative Gain (NDCG) of Top-N recommendation accuracy was increased by 5%-8% on MovieLens datasets and about 1% on Netflix datasets. The experimental results show that the proposed algorithm can effectively improve the Top-N recommendation accuracy.

Key words: matrix factorization, Top-N recommendation, transmutative Singular Value Decomposition (SVD++) algorithm, item popularity, stochastic gradient descent

中图分类号:

TP311.5
TP18

周瑞环, 赵宏宇. 结合物品流行度的列表级矩阵因子分解算法[J]. 计算机应用, 2018, 38(7): 1877-1881.

ZHOU Ruihuan, ZHAO Hongyu. List-wise matrix factorization algorithm with combination of item popularity[J]. Journal of Computer Applications, 2018, 38(7): 1877-1881.

参考文献

[1] 王国霞,刘贺平.个性化推荐系统综述[J].计算机工程与应用,2012,48(7):66-76.(WANG G X, LIU H P. Survey of personalized recommendation system[J]. Computer Engineering and Applications, 2012, 48(7):66-76.)
[2] DESHPANDE M, KARYPIS G. Item-based Top-N recommendation algorithms[J]. ACM Transactions on Information Systems, 2004, 22(1):143-177.
[3] 项亮.推荐系统实践[M].北京:人民邮电出版社,2012:35-41.(XIANG L. Recommended System in Action[M]. Beijing:Posts & Telecom Press, 2012:35-41.)
[4] SEDHAIN S, MENON A K, SANNER S, et al. AutoRec:autoencoders meet collaborative filtering[C]//Proceedings of the 2015 International Conference on World Wide Web. New York:ACM, 2015:111-112.
[5] WU Y, DUBOIS C, ZHENG A X, et al. Collaborative denoising auto-encoders for top-N recommender systems[C]//Proceedings of the 2016 ACM International Conference on Web Search and Data Mining. New York:ACM, 2016:153-162.
[6] HE X, LIAO L, ZHANG H, et al. Neural collaborative filtering[C]//Proceedings of the 2017 International Conference on World Wide Web. New York:ACM, 2017:173-182.
[7] BELL R, KOREN Y. Lessons from the Netflix prize challenge[J]. ACM SIGKDD Explorations Newsletter, 2007, 9(2):75-79.
[8] KOREN Y. Factorization meets the neighborhood:a multifaceted collaborative filtering model[C]//Proceedings of the 2008 ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM, 2008:426-434.
[9] KABBUR S, NING X, KARYPIS G. FISM:factored item similarity models for top-N recommender systems[C]//Proceedings of the 2013 ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York:ACM, 2013:659-667.
[10] LIU N N, YANG Q. EigenRank:a ranking-oriented approach to collaborative filtering[C]//Proceedings of the 2008 International ACM SIGIR Conference on Research and Development in Information Retrieval. New York:ACM, 2008:83-90.
[11] WESTON J, YEE H, WEISS R J. Learning to rank recommendations with the k-order statistic loss[C]//Proceedings of the 2013 ACM Conference on Recommender Systems. New York:ACM, 2013:245-248.
[12] 黄震华,张佳雯,田春岐,等.基于排序学习的推荐算法研究综述[J].软件学报,2016,27(3):691-713.(HUANG Z H, ZHANG J W, TIAN C Q, et al. Survey on learning-to-rank based recommendation algorithms[J]. Journal of Software, 2016, 27(3):691-713.)
[13] RENDLE S, FREUDENTHALER C, GANTNER Z, et al. BPR:Bayesian personalized ranking from implicit feedback[C]//Proceedings of the 2009 Conference on Uncertainty in Artificial Intelligence. Oregon:AUAI Press, 2009:452-461.
[14] LIU J, WU C, XIONG Y, et al. List-wise probabilistic matrix factorization for recommendation[J]. Information Sciences, 2014, 278:434-447.
[15] WEIMER M, KARATZOGLOU A, LE Q V, et al. COFI RANK-maximum margin matrix factorization for collaborative ranking[C]//Proceedings of the 2007 International Conference on Neural Information Processing Systems. Berlin:Springer, 2007:1593-1600.
[16] CAO Z, QIN T, LIU T Y, et al. Learning to rank:from pairwise approach to listwise approach[C]//Proceedings of the 2007 International Conference on Machine Learning. New York:ACM, 2007:129-136.
[17] SHI Y, LARSON M, HANJALIC A. List-wise learning to rank with matrix factorization for collaborative filtering[C]//Proceedings of the 2010 ACM Conference on Recommender Systems. New York:ACM, 2010:269-272.
[18] 赵向宇.Top-N协同过滤推荐技术研究[D].北京:北京理工大学,2014:21-33.(ZHAO X Y. Research on Top-N recommendation with collaborative filtering[D]. Beijing:Beijing Institute of Technology, 2014:21-33.) ZHOU Ruihuan, born in 1993, M. S. candidate. His research interests include personalized recommendation algorithm, machine learning.ZHAO Hongyu, born in 1971, Ph. D., associate professor. His research interests include pattern recognition, artificial intelligence, information theory and coding.

结合物品流行度的列表级矩阵因子分解算法

List-wise matrix factorization algorithm with combination of item popularity

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics