基于非负矩阵分解的Slope One算法
|
摘要
针对协同过滤推荐算法中Slope One算法在稀疏数据集中推荐精度低的问题,利用矩阵分解在解决矩阵稀疏性方面的优势,将非负矩阵分解技术引入到用户-项目评分矩阵的降维处理中,将原有的稀疏评分矩阵进行非负分解,改善了矩阵的稀疏性,优化Slope One算法.从实验数据可以看出,与原始的CF算法进行比较,NMF-Slope One算法有较好的推荐效果.在数据稀疏的条件下,确定参数进行实验.实验结果表明,该方法提高了Slope One算法在数据稀疏下的精度和推荐质量.
The good performance of matrix decomposition in solving matrix sparsity was used in order to solve the problem that the Slope One algorithm has low recommendation accuracy in the sparse data set in the collaborative filtering recommendation algorithm. The nonnegative matrix factorization technology was introduced into the dimension reduction of the user-item rating matrix in order to optimize the Slope One algorithm. The original sparse scoring matrix was non-negatively decomposed in order to improve the sparsity of the matrix. The experimental results show that the NMF-Slope One algorithm has a good recommendation effect compared with the original CF algorithm. Parameters were determined for experimentation under conditions of sparse data. The proposed method improves the accuracy and the recommendation quality of the Slope One algorithm under data sparseness.
The good performance of matrix decomposition in solving matrix sparsity was used in order to solve the problem that the Slope One algorithm has low recommendation accuracy in the sparse data set in the collaborative filtering recommendation algorithm. The nonnegative matrix factorization technology was introduced into the dimension reduction of the user-item rating matrix in order to optimize the Slope One algorithm. The original sparse scoring matrix was non-negatively decomposed in order to improve the sparsity of the matrix. The experimental results show that the NMF-Slope One algorithm has a good recommendation effect compared with the original CF algorithm. Parameters were determined for experimentation under conditions of sparse data. The proposed method improves the accuracy and the recommendation quality of the Slope One algorithm under data sparseness.
引文
[1]CACHEDA F,FORMOSO V,FERNáNDEZ D,et al.Comparison of collaborative filtering algorithms:Limitations of current techniques and proposals for scalable,high-performance recommender systems[J].ACM Transactions on the Web,2011,5(1):1-33.
[2]LEMIRE D,MACLACHLAN A.Slope One predictors for online rating-based collaborative filtering[C]//Proceedings of the 2005 SIAM International Conference on Data Mining.Philadelphia:Society for Industrial and Applied Mathematics Publications,2005:471-475.
[3]KARYDI E,MARGARITIS K G.Multithreaded implementation of the Slope One algorithm for collaborative filtering[C]//8th IFIP WG 12.5International Conference on Artificial Intelligence Applications and Innovations.New York:Springer,2012:117-125.
[4]KOREN Y,BELL R.Advances in collaborative filtering[M].New York:Springer,2015:77-118.
[5]TIAN S,OU L.An improved Slope One algorithm combining KNN method weighted by user similarity[C]//17th International Conference on Web-Age Information Management.Berlin:Springer,2016:88-98.
[6]BASU A,VAIDYA J,KIKUCHI H.Perturbation based privacy preserving Slope One predictors for collaborative filtering[C]//6th IFIP WG 11.11International Conference on Trust Management VI.New York:Springer,2012:17-35.
[7]MAO C,CHEN J.QoS prediction for web services based on similarity-aware Slope One collaborative filtering[J].Informatica(Slovenia),2013,37(2):139-148.
[8]LIU Y,LIU D,XIE H,et al.A research on the improved slope one algorithm for collaborative filtering[J].International Journal of Computing Science and Mathematics,2016,7(3):245-253.
[9]SAEED M,MANSOORI E G.A new slope one based recommendation algorithm using virtual predictive items[J].Journal of Intelligent Information Systems,2018,50(3):527-547.
[10]WANG Q X,LUO X,LI Y,et al.Incremental Slopeone recommenders[J].Neurocomputing,2018,272:606-618.
[11]LEE D D,SEUNG H S.Learning the parts of objects with non-negative matrix factorization[J].Nature,1999,401(6755):788-791.
[12]ZONG L,ZHANG X,ZHAO L,et al.Multi-view clustering via multi-manifold regularized nonnegative matrix factorization[J].Neural Networks,2017,88:74-89.
[13]LU S,HONG M,WANG Z.A Nonconvex splitting method for symmetric nonnegative matrix factorization:convergence analysis and optimality[J].IEEE Transactions on Signal Processing,2017,65(12):3120-3135.
[14]ALQUIER P,GUEDJ B.An oracle inequality for quasi-Bayesian nonnegative matrix factorization[J].Mathematical Methods of Statistics,2017,26(1):55-67.
[15]HARPER F M,KONSTAN J A.The MovieLens datasets:history and context[J].ACMTransactions on Interactive Intelligent Systems,2015,5(4):1-19.
[16]林德军,孟祥武.基于奇异值分解的Slope One算法[J].新型工业化,2012,2(11):12-17.LIN De-jun,MENG Xiang-wu.Slope One algorithm based on single value decomposition[J].The Journal of New Industrialization,2012,2(11):12-17.
[2]LEMIRE D,MACLACHLAN A.Slope One predictors for online rating-based collaborative filtering[C]//Proceedings of the 2005 SIAM International Conference on Data Mining.Philadelphia:Society for Industrial and Applied Mathematics Publications,2005:471-475.
[3]KARYDI E,MARGARITIS K G.Multithreaded implementation of the Slope One algorithm for collaborative filtering[C]//8th IFIP WG 12.5International Conference on Artificial Intelligence Applications and Innovations.New York:Springer,2012:117-125.
[4]KOREN Y,BELL R.Advances in collaborative filtering[M].New York:Springer,2015:77-118.
[5]TIAN S,OU L.An improved Slope One algorithm combining KNN method weighted by user similarity[C]//17th International Conference on Web-Age Information Management.Berlin:Springer,2016:88-98.
[6]BASU A,VAIDYA J,KIKUCHI H.Perturbation based privacy preserving Slope One predictors for collaborative filtering[C]//6th IFIP WG 11.11International Conference on Trust Management VI.New York:Springer,2012:17-35.
[7]MAO C,CHEN J.QoS prediction for web services based on similarity-aware Slope One collaborative filtering[J].Informatica(Slovenia),2013,37(2):139-148.
[8]LIU Y,LIU D,XIE H,et al.A research on the improved slope one algorithm for collaborative filtering[J].International Journal of Computing Science and Mathematics,2016,7(3):245-253.
[9]SAEED M,MANSOORI E G.A new slope one based recommendation algorithm using virtual predictive items[J].Journal of Intelligent Information Systems,2018,50(3):527-547.
[10]WANG Q X,LUO X,LI Y,et al.Incremental Slopeone recommenders[J].Neurocomputing,2018,272:606-618.
[11]LEE D D,SEUNG H S.Learning the parts of objects with non-negative matrix factorization[J].Nature,1999,401(6755):788-791.
[12]ZONG L,ZHANG X,ZHAO L,et al.Multi-view clustering via multi-manifold regularized nonnegative matrix factorization[J].Neural Networks,2017,88:74-89.
[13]LU S,HONG M,WANG Z.A Nonconvex splitting method for symmetric nonnegative matrix factorization:convergence analysis and optimality[J].IEEE Transactions on Signal Processing,2017,65(12):3120-3135.
[14]ALQUIER P,GUEDJ B.An oracle inequality for quasi-Bayesian nonnegative matrix factorization[J].Mathematical Methods of Statistics,2017,26(1):55-67.
[15]HARPER F M,KONSTAN J A.The MovieLens datasets:history and context[J].ACMTransactions on Interactive Intelligent Systems,2015,5(4):1-19.
[16]林德军,孟祥武.基于奇异值分解的Slope One算法[J].新型工业化,2012,2(11):12-17.LIN De-jun,MENG Xiang-wu.Slope One algorithm based on single value decomposition[J].The Journal of New Industrialization,2012,2(11):12-17.