Please wait a minute...
浙江大学学报(工学版)
Journal of ZheJiang University (Engineering Science)  2019, Vol. 53 Issue (3): 548-554    DOI: 10.3785/j.issn.1008-973X.2019.03.016
Computer Technology     
Heterogeneous information network representation learning based on transition probability matrix (HINtpm)
Ting-ting ZHAO(),zhe WANG*(),Yi-nan LU
College of Computer Science and Technology, Key Laboratory of Symbolic Computation and Knowledge Engineering, Ministry of Education, Jilin University, Changchun 130012, China
Download: HTML     PDF(860KB) HTML
Export: BibTeX | EndNote (RIS)      

Abstract  

First, the final probability transition matrix was obtained by combining the first-order and second-order similarity of the nodes, according to the meta-path and the commuting matrix. Then, a Denoisin Auto-encoder was used to reduce the dimension of probability transition matrix for getting the node representation in heterogeneous information network. Finally, the node representation in heterogeneous information network was classified by gradient boosting decision tree (GBDT) and the classification accuracy under different percentage training set was obtained. Use the clustering index normalized mutual information (NMI) to evaluate the clustering effect and use T-SNE to show the visual effect. Experiments were performed on data sets DBLP and AMiner. The proposed heterogeneous information network representation learning based on transition probability matrix (HINtpm) was compared with DeepWalk, node2vec and metapath2vec methods. As results, compared with DeepWalk method, HINtpm improved the classification accuracy by 24% the maximum on the application task-node classify and increased the clustering index NMI by 13% the maximum.

Key words： network representation learning      heterogeneous information network (HIN)      transition probability matrix      meta-path      nodes' similarity      auto-encoder     
Received: 05 February 2018      Published: 04 March 2019
CLC:  TP 391  
Corresponding Authors: zhe WANG     E-mail: zhaott16@mails.jlu.edu.cn;wz2000@jlu.edu.cn
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Ting-ting ZHAO
zhe WANG
Yi-nan LU
Cite this article:

Ting-ting ZHAO,zhe WANG,Yi-nan LU. Heterogeneous information network representation learning based on transition probability matrix (HINtpm). Journal of ZheJiang University (Engineering Science), 2019, 53(3): 548-554.

URL:

http://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2019.03.016     OR     http://www.zjujournals.com/eng/Y2019/V53/I3/548


基于传播概率矩阵的异构信息网络表示学习

根据元路径和可交换矩阵，结合节点一阶和二阶相似性得到最后的传播概率矩阵；利用降噪自动编码器对传播概率矩阵进行降维得到异构信息网络的节点表示；将异构信息网络的节点表示用梯度提升树（GBDT）分类，得到不同百分比训练集下的分类准确率，用聚类指标标准化互信息（NMI）评价聚类效果，用T-SNE展现可视化效果. 在数据集DBLP和AMiner上分别进行实验，相比DeepWalk、node2vec和metapath2vec方法，在应用任务节点分类上，所提出的基于传播概率矩阵的异构信息网络表示学习（HINtpm）的准确率与DeepWalk相比最高提升了24%，聚类指标NMI与DeepWalk相比最高提升了13%.


关键词： 网络表示学习,  异构信息网络（HIN）,  传播概率矩阵,  元路径,  节点相似性,  自动编码器 
Fig.1 Example of typical heterogeneous information network (HIN): DBLP (digital bibliography & library project)
Fig.2 Meta schema of dataset DBLP
Fig.3 Example of DBLP meta path: A-P-C-P-A
Fig.4 First-order and second-order similarity of homogeneous information network
Fig.5 Model for denoising auto-encoder
算法 R=10% R=20% R=30% R=40% R=50% R=60% R=70% R=80% R=90%
DeepWalk 0.655 0 0.713 9 0.754 8 0.743 9 0.780 7 0.768 0 0.788 4 0.795 6 0.786 2
Node2vec 0.653 5 0.735 5 0.764 4 0.772 7 0.783 1 0.761 7 0.781 5 0.786 2 0.817 6
Pathsim 0.790 2 0.816 8 0.827 5 0.826 7 0.824 5 0.833 1 0.833 4 0.830 2 0.888 1
Metapath2vec 0.866 2 0.896 9 0.888 4 0.896 0 0.897 6 0.907 1 0.911 8 0.920 1 0.880 5
HINtpm 0.893 6 0.914 3 0.926 7 0.929 8 0.929 9 0.937 1 0.949 6 0.944 3 0.941 5
HINtpm0 0.897 3 0.925 3 0.935 3 0.926 1 0.938 1 0.945 0 0.945 2 0.939 0 0.968 6
Tab.1 Classification accuracy of different algorithms on dataset DBLP
算法 R=10% R=20% R=30% R=40% R=50% R=60% R=70% R=80% R=90%
DeepWalk 0.858 4 0.873 0 0.888 5 0.898 4 0.907 4 0.910 6 0.915 2 0.910 0 0.902 5
Node2vec 0.888 6 0.914 6 0.923 0 0.930 4 0.940 4 0.944 6 0.943 7 0.932 8 0.935 0
pathsim 0.954 4 0.965 8 0.997 17 0.971 5 0.972 1 0.974 8 0.979 2 0.980 0 0.982 5
Metapath2vec 0.898 5 0.928 5 0.935 2 0.937 6 0.935 2 0.939 9 0.940 5 0.953 9 0.944 0
HINtpm 0.947 5 0.961 4 0.960 3 0.963 7 0.961 0 0.963 8 0.969 7 0.975 4 0.984 3
HINtpm0 0.963 6 0.967 6 0.969 7 0.976 8 0.976 7 0.975 6 0.983 5 0.985 9 0.985 0
Tab.2 Classification accuracy of different algorithms on dataset AMiner
算法 NMI-DBLP NMI- AMiner
DeepWalk 0.483 4 0.781 8
Node2vec 0.499 0 0.831 5
Metapath2vec 0.570 7 0.595 8
pathsim 0.334 1 0.600 4
HINtpm 0.612 9 0.831 7
HINtpm0 0.608 6 0.802 0
Tab.3 Normalized mutual information (NMI) comparison of different algorithms on datasets DBLP and AMiner
Fig.6 Visual effect of different algorithms on database AMiner
[1]   涂存超, 杨成, 刘知远, 等 网络表示学习综述[J]. 中国科学: 信息科学, 2017, 47 (8): 980- 996
TU Cun-Chao, YANG Cheng, LIU Zhi-Yuan, at el Network representation learning: an overview[J]. SCIENTIA SINICA: Informations, 2017, 47 (8): 980- 996
[2]   ROWEIS S T, SAUL L K Nonlinear dimensionality reduction by locally linear embedding[J]. Science, 2000, 290 (5500): 2323- 2326
doi: 10.1126/science.290.5500.2323
[3]   COX T F, COX M A A. Multidimensional scaling [M]. Boca Raton: CRC Press, 2000: 123–141.
[4]   BELKIN M, NIYOGI P. Laplacian eigenmaps and spectral techniques for embedding and clustering [C] // Proceedings of the 14th International Conference on Neural Information Processing Systems: Natural and Synthetic. Cambridge: MIT Press, 2001: 585–591.
[5]   CHEN M, YANG Q, TANG X O. Directed graph embedding [C] // Proceedings of the 20th International Joint Conference on Artificial Intelligence. San Francisco: Morgan Kaufmann Publishers Inc, 2007: 2707–2712.
[6]   NALLAPATI R, COHEN W W. Link-PLSA-LDA: a new unsupervised model for topics and influence of blogs [C] // Proceedings of the 2nd International Conference on Weblogs and Social Media. Seattle: AAAI, 2008: 84–92.
[7]   CHANG J, BLEI D M. Relational topic models for document networks [C] // Proceedings of the 12th International Conference on Artificial Intelligence and Statistics. Cambridge: JMLR, 2009: 81–88.
[8]   LE T M V, LAUW H W. Probabilistic latent document network embedding [C] // Proceedings of the 2014 International Conference on Data Mining. Washington: IEEE Computer Society, 2014: 270–279.
[9]   WARD C K Word2Vec[J]. Natural Language Engineering, 2016, 23 (1): 155- 162
doi: 10.1017/S1351324916000334
[10]   TANG J, QU M, WANG M, et al. Line: large-scale information network embedding [C] // Proceedings of the 24th International Conference on World Wide Web. Florence: WWW, 2015: 1067–1077
[11]   WANG D, CUI P, ZHU W. Structural deep network embedding [C] // Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco: ACM, 2016: 1225–1234
[12]   TANG J, QU M, MEI Q. PTE: predictive text embedding through large-scale heterogeneous text networks [C] // ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Sydney: ACM, 2015: 1165–1174.
[13]   CAO S S, LU W, XU Q K. GraRep: learning graph representations with global structural information [C] // ACM International on Conference on Information and Knowledge Management. Melbourne: ACM, 2015: 891–900.
[14]   YANG J, LESKOVEC J. Overlapping community detection at scale: a nonnegative matrix factorization approach [C] // Proceedings of the 6th ACM International Conference on Web Search and Data Mining. Rome: ACM, 2013: 587–596
[15]   LEY M. The DBLP Computer science bibliography: evolution, research issues, perspectives [C] // String Processing and Information Retrieval, International Symposium. Portugal: SPIRE, 2002: 1–10.
[16]   TANG J, ZHANG J, YAO L M, et al. ArnetMiner: extraction and mining of academic social networks [C] // ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Las Vegas: SIGKDD, 2008: 990–998.
[17]   CHANG S Y, HAN W, TANG J L, et al. Heterogeneous network embedding via deep architectures [C] // ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Sydney: ACM, 2015: 119–128.
[18]   JI M, HAN J W, DANILEVSKY M. Ranking-based classification of heterogeneous information networks [C] // ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Diego: ACM, 2011: 1298–1306.
[19]   DONG Y, CHAWLA N V, SWAMI A. Metapath2vec: Scalable representation learning for heterogeneous networks [C] // Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Nova Scotia: ACM, 2017: 135–144.
[20]   PEROZZI B, ALRFOU R, SKIENA S. DeepWalk: online learning of social representations [C]// ACM Sigkdd International Conference on Knowledge Discovery & Data Mining. New York: SIGKDD, 2014: 701-710.
[21]   GROVER A, LESKOVEC J. node2vec: scalable feature learning for networks [C] // ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco: SIGKDD, 2016: 855–864.
[1] Xue-feng ZHANG,Xiu-li CHEN,De-wen SENG. Top-N recommendation algorithm combining user trust and influence[J]. Journal of ZheJiang University (Engineering Science), 2020, 54(2): 311-319.
[2] Wan-liang WANG,Xiao-han YANG,Yan-wei ZHAO,Nan GAO,Chuang LV,Zhao-juan ZHANG. Image enhancement algorithm with convolutional auto-encoder network[J]. Journal of ZheJiang University (Engineering Science), 2019, 53(9): 1728-1740.