基于先验知识的非负矩阵半可解释三因子分解算法

doi:10.11772/j.issn.1001-9081.2021040927

《计算机应用》唯一官方网站 ›› 2022, Vol. 42 ›› Issue (3): 671-675.DOI: 10.11772/j.issn.1001-9081.2021040927

• 2021年中国计算机学会人工智能会议(CCFAI 2021) • 上一篇下一篇

基于先验知识的非负矩阵半可解释三因子分解算法

陈露¹^,², 张晓霞¹^,²(), 于洪¹^,²

^1.计算智能重庆市重点实验室（重庆邮电大学），重庆 400065
^2.重庆邮电大学计算机科学与技术学院，重庆 400065

收稿日期:2021-06-03 修回日期:2021-08-24 接受日期:2021-08-25 发布日期:2021-11-09 出版日期:2022-03-10
通讯作者: 张晓霞
作者简介:陈露（1997—），男，重庆人，硕士研究生，主要研究方向：三支决策、粒计算、推荐系统、数据挖掘、认知计算
于洪（1980—），女，重庆人，教授，博士，CCF会员，主要研究方向：工业大数据分析与处理、智能决策、知识发现、粒计算、三支聚类、智能推荐。
基金资助:
国家重点研发计划项目(2019YFB2103000);国家自然科学基金资助项目(61936001);重庆市自然科学基金资助项目(cstc2020jcyj?msxmX0737);重庆市教委科学技术研究青年项目(KJQN201900638)

Partially explainable non-negative matrix tri-factorization algorithm based on prior knowledge

Lu CHEN¹^,², Xiaoxia ZHANG¹^,²(), Hong YU¹^,²

^1.Chongqing Key Laboratory of Computational Intelligence （Chongqing University of Posts and Telecommunications），Chongqing 400065，China
^2.College of Computer Science and Technology，Chongqing University of Posts and Telecommunications，Chongqing 400065，China

Received:2021-06-03 Revised:2021-08-24 Accepted:2021-08-25 Online:2021-11-09 Published:2022-03-10
Contact: Xiaoxia ZHANG
About author:CHEN Lu， born in 1997， M. S. candidate. His research interests include three-way decision， granular computing， recommender system， data mining， cognitive computing.
YU Hong， born in 1980， Ph. D.， professor. Her research interests include industrial big data analysis and processing， intelligent decision， knowledge discovery， granular computing， three-way clustering， intelligent recommendation.
Supported by:
National Key Research and Development Program of China(2019YFB2103000);National Natural Science Foundation of China(61936001);Natural Science Foundation of Chongqing(cstc2020jcyj-msxmX0737);Youth Science and Technology Research Program of Chongqing Municipal Education Commission(KJQN201900638)

摘要/Abstract

摘要：

非负矩阵三因子分解是潜在因子模型中的重要组成部分，由于能将原始数据矩阵分解为三个相互约束的潜因子矩阵，被广泛应用于推荐系统、迁移学习等研究领域，但目前还没有非负矩阵三因子分解的可解释性方面的研究工作。鉴于此，将用户评论文本信息当作先验知识，设计了一种基于先验知识的非负矩阵半可解释三因子分解（PE-NMTF）算法。首先利用情感分析技术提取用户评论文本信息的情感极性偏好；然后更改了非负矩阵三因子分解算法的目标函数和更新公式，巧妙地将先验知识嵌入到算法中；最后在推荐系统冷启动任务的Yelp和Amazon数据集以及图像零次识别任务的AwA和CUB数据集上与非负矩阵分解、非负矩阵三因子分解算法做了大量对比实验，实验结果表明所提算法在均方根误差（RMSE）、归一化折损累计增益（NDCG）、归一化互信息（NMI）和准确率（ACC）上都表现优异，且利用先验知识进行非负矩阵三因子分解的解释具有可行性和有效性。

Abstract:

Non-negative Matrix Tri-Factorization （NMTF） is an important part of the latent factor model. Because this algorithm decomposes the original data matrix into three mutually constrained latent factor matrices， it has been widely used in research fields such as recommender systems and transfer learning. However， there is no research work on the interpretability of non-negative matrix tri-factorization. From this view， by regarding the user comment text information as prior knowledge， Partially Explainable Non-negative Matrix Tri-Factorization （PE-NMTF） algorithm was designed based on prior knowledge. Firstly， sentiment analysis technology was used by to extract the emotional polarity preferences of user comment text information. Then， the objective function and updating formula in non-negative matrix tri-factorization algorithm were changed， embedding prior knowledge into the algorithm. Finally， a large number of experiments were carried out on the Yelp and Amazon datasets for the cold start task of the recommender system and the AwA and CUB datasets for the image zero-shot task to compare the proposed algorithm with the non-negative matrix factorization and the non-negative matrix three-factor decomposition algorithms. The experimental results show that the proposed algorithm performs well on RMSE （Root Mean Square Error）， NDCG （Normalized Discounted Cumulative Gain）， NMI （Normalized Mutual Information）， and ACC （ACCuracy）， and the feasibility and effectiveness of the non-negative matrix tri-factorization were verified by using prior knowledge.

Key words: non-negative matrix tri-factorization, recommender system, interpretable machine learning, prior knowledge, latent factor model

中图分类号:

TP181

陈露, 张晓霞, 于洪. 基于先验知识的非负矩阵半可解释三因子分解算法[J]. 计算机应用, 2022, 42(3): 671-675.

Lu CHEN, Xiaoxia ZHANG, Hong YU. Partially explainable non-negative matrix tri-factorization algorithm based on prior knowledge[J]. Journal of Computer Applications, 2022, 42(3): 671-675.

图/表 4

图1 非负矩阵三因子分解示例

Fig. 1 Example of non-negative matrix tri-factorization

表1 四种算法在冷启动任务上的实验结果

Tab. 1 Experimental results of four algorithms on cold-start tasks

数据集	NMF		NMTF		FE-NMTF		PF-NMTF
数据集	RMSE	NDCG	RMSE	NDCG	RMSE	NDCG	RMSE	NDCG
Yelp	1.040 9	0.838 1	1.054 7	0.837 3	1.023 6	0.843 0	1.027 4	0.843 0
Amazon	0.997 7	0.386 3	1.030 1	0.386 5	0.983 4	0.386 9	0.991 1	0.386 8

表2 四种算法在零次识别上的实验结果

Tab. 2 Experimental results of four algorithms on zero-shot tasks

数据集	NMF		NMTF		FE-NMTF		PF-NMTF
数据集	NMI	ACC	NMI	ACC	NMI	ACC	NMI	ACC
AwA	0.8333	0.6322	0.8364	0.6332	0.8676	0.6856	0.8553	0.6966
CUB	0.7265	0.2834	0.7401	0.3029	0.7575	0.4009	0.7449	0.3837

图2 四种算法在4个数据集的收敛性

Fig. 2 Convergence of four algorithms on 4 datasets

参考文献 27

1	FAYYAD， USAMA M. Advances in knowledge discovery and data mining ［C］// Proceedings of the 15th Pacific-Asia Conference on Knowledge Discovery and Data Mining. New York： ACM， 2011： 113-121. 10.1007/978-3-642-20847-8
2	YANG X， HUANG K， ZHANG R， et al. Learning latent features with infinite nonnegative binary matrix tri factorization［J］. IEEE Transactions on Emerging Topics in Computational Intelligence， 2018， 2（6）： 450-463. 10.1109/tetci.2018.2806934
3	LONG B， ZHANG M Z， YU P S. Co-clustering by block value decomposition ［C］// Proceedings of the 11th ACM SIGKDD International Conference on Knowledge Discovery in Data Mining. New York： ACM， 2005： 40-53. 10.1145/1081870.1081949
4	WANG H， NIE F， HUANG H， et al. Fast non-negative matrix tri-factorization for large-scale data co-clustering ［C］// Proceedings of the 22nd International Joint Conference on Artificial Intelligence. Menlo Park， CA： AAAI， 2011： 558-569. 10.1109/icdm.2011.109
5	TIAN J， QU Y. A novel framework for top-n recommendation based on non-negative matrix tri-factorization［C］// Proceedings of the 24th International Conference on Industrial Engineering and Engineering Management. Cham： Springer， 2019： 233-241. 10.1007/978-981-13-3402-3_36
6	LEE D D， SEUNG H F， Learning the parts of objects by non-negative matrix factorization ［J］. Nature， 1999， 401（2）： 788-791. 10.1038/44565
7	HUANG S， XU Z， LYU J. Adaptive local structure learning for document co-clustering ［J］. Knowledge-Based Systems， 2018， 81（32） 148：7484. 10.1016/j.knosys.2018.02.020
8	LI T， ZHANG Y， SINDHWANI V. A non-negative matrix tri-factorization approach to sentiment classification with lexical prior knowledge［C］// Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing. Stroudsburg， PA： Association for Computational Linguistics， 2009： 244-252. 10.3115/1687878.1687914
9	RAMOS J. Using TF-IDF to determine word relevance in document queries［C］// Proceedings of the 1st Instructional Conference on Machine Learning. Stroudsburg， PA： Association for Computational Linguistics， 2003： 29-48.
10	FU G， WANG J， DOMENICONI C， YU G. Matrix factorization-based data fusion for the prediction of lncRNA-disease associations ［J］. Bioinformatics， 2017， 34（9）： 1529-1537. 10.1093/bioinformatics/btx794
11	CEDDIA G， PINOLI P， CERI S， et al. Non-negative matrix tri-factorization for data integration and network-based drug repositioning［C］// Proceedings of the 2019 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology （CIBCB）. Piscataway： IEEE， 2019： 992-1003. 10.1109/cibcb.2019.8791474
12	高科. 基于隐空间的子空间学习［D］. 天津：天津大学， 2019： 21.
	GAO K. Subspace learning based on latent space［D］. Tianjing： Tianjing University， 2019： 21.
13	汪涛，刘阳，席耀一. 基于图正则化非负矩阵分解的二分网络社区发现算法［J］. 电子与信息学报， 2015， 37（9）： 2238-2245. 10.11999/JEIT141649
	WANG T， LIU Y， XI Y Y. Identifying community in bipartite networks using graph regularized-based non-negative matrix factorization［J］. Journal of Electronics & Information Technology， 2015， 37（9）： 2238-2245. 10.11999/JEIT141649
14	金弟，何静. 基于非负矩阵三因子分解的属性网络半监督社团发现的方法： CN110851732A［P］. 2020-02-28. 10.1007/s11704-020-9203-0
	JIN D， HE J. Semi-supervised community discovery method for attribute networks based on non-negative matrix tri-factorization： CN110851732A［P］. 2020-02-28. 10.1007/s11704-020-9203-0
15	TAN B， SONG Y， ZHONG E， et al. Transitive transfer learning ［C］// Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2019： 1155-1164.
16	JUNIOR W， PERES S M， FREIRE V， et al. OvNMTF algorithm： an overlapping non-negative matrix tri-factorization for coclustering ［C］// Proceedings of the 2020 International Joint Conference on Neural Networks . Piscataway： IEEE， 2020： 348-353. 10.1109/ijcnn48605.2020.9207364
17	BUONO N DEL， PIO G. Non-negative matrix tri-factorization for co-clustering： an analysis of the block matrix［J］. Information Sciences， 2015， 301： 13-26. 10.1016/j.ins.2014.12.058
18	OPAR A， ZUPAN B， ZITNIK M. Fast optimization of non-negative matrix tri-factorization ［J］. PLoS One， 2019， 14（9）： 12-21. 10.1371/journal.pone.0217994
19	ABDOLLAHI B， NASRAOUI O. Using explainability for constrained matrix factorization ［C］// Proceedings of the 11th ACM Conference on Recommender Systems. New York： ACM， 2017： 79-83. 10.1145/3109859.3109913
20	LU Y， CASTELLANOS M， DAYAL U， et al. Automatic construction of a context-aware sentiment lexicon： an optimization approach［C］// Proceedings of the 20th International Conference on World Wide Web. Cham： Springer， 2011： 221-231. 10.1145/1963405.1963456
21	ZHANG Y， LAI G， ZHANG M， et al. Explicit factor models for explainable recommendation based on phrase-level sentiment analysis［C］// Proceedings of the 37th International ACM SIGIR Conference. New York： ACM， 2014： 83-92. 10.1145/2600428.2609579
22	TAO Y， JIA Y， WANG N， et al. The FacT： taming latent factor models for explainability with factorization trees［C］// Proceedings of the 42nd International ACM SIGIR Conference. New York： ACM， 2019： 413-432. 10.1145/3331184.3331244
23	WU Q， TAN M， LI X， et al. NMFE-SSCC： non-negative matrix factorization ensemble for semi-supervised collective classification ［J］. Knowledge-Based Systems， 2015， 89（15）： 160-172. 10.1016/j.knosys.2015.06.026
24	XIAN Y， LAMPERT C H， SCHIELE B， et al. Zero-shot learning - a comprehensive evaluation of the good， the bad and the ugly ［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2018， 40（8）： 182-199. 10.1109/cvpr.2018.00581
25	KEMP C， TENENBAUM J B， GRIFFITHS T L， et al. Learning systems of concepts with an infinite relational model ［J］. Cognitive Science， 2006， 12（21）： 313-329.
26	WAH C， BRANSON S， WELINDER P， et al. Caltech-UCSD birds- 200-2011， CNS-TR-2010-001［R/OL］. California Institute of Technology，2010［2021-06-01］..
27	CHEN G， WANG F， ZHANG C. Collaborative filtering using orthogonal nonnegative matrix tri-factorization［J］. Information Processing & Management， 2009， 45（3）： 113-125. 10.1016/j.ipm.2008.12.004

[1]	蒋雯静, 熊熙, 李中志, 李斌勇. 基于无采样协作知识图网络的推荐系统[J]. 《计算机应用》唯一官方网站, 2022, 42(4): 1057-1064.
[2]	顾军华, 王锐, 李宁宁, 张素琪. 融合协同过滤信息的知识图注意力网络[J]. 《计算机应用》唯一官方网站, 2022, 42(4): 1087-1092.
[3]	刘忠慧, 王梓宥, 闵帆. 近似概念的遗传生成算法及其推荐应用[J]. 《计算机应用》唯一官方网站, 2022, 42(2): 412-418.
[4]	潘仁志, 钱付兰, 赵姝, 张燕平. 基于卷积神经网络交互的用户属性偏好建模的推荐模型[J]. 《计算机应用》唯一官方网站, 2022, 42(2): 404-411.
[5]	刘欢, 李晓戈, 胡立坤, 胡飞雄, 王鹏华. 基于知识图谱驱动的图神经网络推荐模型[J]. 计算机应用, 2021, 41(7): 1865-1870.
[6]	林怿星, 唐华. 基于异构信息网络的混合推荐模型[J]. 计算机应用, 2021, 41(5): 1348-1355.
[7]	骆锦潍, 刘杜钢, 潘微科, 明仲. 基于改进的倾向得分估计的无偏推荐模型[J]. 《计算机应用》唯一官方网站, 2021, 41(12): 3508-3514.
[8]	李改, 李磊, 张佳强. 融合显/隐式反馈的社会化协同排序推荐算法[J]. 《计算机应用》唯一官方网站, 2021, 41(12): 3515-3520.
[9]	陈宪聪, 潘微科, 明仲. 面向异构单类协同过滤的阶段式变分自编码器[J]. 《计算机应用》唯一官方网站, 2021, 41(12): 3499-3507.
[10]	高铭蔚, 桑楠, 杨茂林. 基于胶囊网络的交互式网络电视视频点播推荐模型[J]. 《计算机应用》唯一官方网站, 2021, 41(11): 3171-3177.
[11]	任柯舟, 彭甫镕, 郭鑫, 王喆, 张晓静. 动态融合社交信息的社会化推荐[J]. 计算机应用, 2021, 41(10): 2806-2812.
[12]	李翔锟, 贾彩燕. 融合重叠社区正则化及隐式反馈的协同过滤方法[J]. 计算机应用, 2021, 41(1): 53-59.
[13]	李春英, 汤庸, 肖政宏, 李天送. 学术社交网络中的权威学者推荐模型[J]. 计算机应用, 2020, 40(9): 2594-2599.
[14]	秦婧, 张青博, 王斌. 关注长尾物品的推荐方法[J]. 《计算机应用》唯一官方网站, 2020, 40(2): 454-458.
[15]	许长青, 陈振杰, 侯仁福. 融合不精准先验知识的Landsat 8 OLI影像深度学习分类方法[J]. 计算机应用, 2020, 40(12): 3550-3557.

基于先验知识的非负矩阵半可解释三因子分解算法

Partially explainable non-negative matrix tri-factorization algorithm based on prior knowledge

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 27

相关文章 15

编辑推荐

Metrics