基于多层感知机的个性化链接排序预测
|
摘要
社交网络链接预测,即通过历史社交网络结构信息,预测未来一段时间内社交用户之间可能会产生新的链接关系,是社会网络分析中的一个重要问题。现有的模型挖掘了用户之间的浅层交互关系,或者通过深层网络去学习用户的特征描述。然而,由于社会网络数据极其稀疏,现有的模型在链接预测的表现上存在一定的提升空间。针对上述问题,文章提出基于多层感知机的个性化链接排序预测模型(PRMLP),从而实现了社交链接预测任务。PRMLP同时考虑了用户之间的交互关系,并采用了多层网络结构深入挖掘社会网络的拓扑结构,因此能够学习得到更精准的用户特征描述。文章针对模型训练中正负样本不平衡的问题提出了解决方案,在2个真实数据集进行的实验表明,文中提出的基于多层感知机的个性化链接排序预测模型相对于现有的其他链接预测模型表现更优。
Link prediction is the task of predicting the possible link relationships between users based on the current social network structure, which plays an important role in social network analysis. The current solutions either model shallow interaction relationships between users or simply adopt deep learning models for the prediction task. Nevertheless, due to the sparsity of the social network data, the performance of these methods is not satisfactory. In this paper, a new model of personalized link prediction based on multilayer perception named PRMLP is proposed. Specifically, to deal with the data sparsity problem, PRMLP model adopts a multilayer perception neural network that considers both the complementary advantage of the shallow user interaction relationship and the deep network structure for learning useful user representations. Furthermore, a solution to deal with the data imbalance problem in model learning process that there are much more negative links than the positive links in the social network is proposed. The experiments on two real datasets demonstrate that the proposed method outperforms the state-of-the-art methods.
Link prediction is the task of predicting the possible link relationships between users based on the current social network structure, which plays an important role in social network analysis. The current solutions either model shallow interaction relationships between users or simply adopt deep learning models for the prediction task. Nevertheless, due to the sparsity of the social network data, the performance of these methods is not satisfactory. In this paper, a new model of personalized link prediction based on multilayer perception named PRMLP is proposed. Specifically, to deal with the data sparsity problem, PRMLP model adopts a multilayer perception neural network that considers both the complementary advantage of the shallow user interaction relationship and the deep network structure for learning useful user representations. Furthermore, a solution to deal with the data imbalance problem in model learning process that there are much more negative links than the positive links in the social network is proposed. The experiments on two real datasets demonstrate that the proposed method outperforms the state-of-the-art methods.
引文
[1] LU L Y,ZHOU T.Link prediction in complex networks:a survery[EB/OL].[2010-10-04].http://arxiv.org/abs/1010.0725.
[2] 张宗宇.社会化网络的链接预测[D].北京:北京邮电大学,2011.
[3] O′MADADHAIN J,HUTCHINS J,SMYTH P.Prediction and ranking algorithms for event-based network data[M].New York:ACM Press,2005:23-30.
[4] LU L Y,JIN C H,ZHOU T.Similarity index based on local paths for link prediction of complex networks[J].Physical Review E,2009,80(1):046122.
[5] LIU W P,LU L Y.Link prediction based on local random walk[J].A letters Journal Exploring the Frontiers of Physics,2010,89(5):58007-58012.
[6] 孙诚,王志海.社会网络中基于神经网络的链路预测方法[J].数学建模及其应用,2017,4(6):10-17.
[7] FEI T,XIA Y X,ZHU B Y.Link prediction in complex networks:a mutual information Perspective[J].PloS ONE,2014,9(9):e107056.
[8] ZHU B Y,XIA Y X.Link prediction in weighted networks:a weighted mutmal information model[J].PLoS ONE,2016,11(2):e0148265.
[9] LU Y D,GUO Y F,KNORHONEN A.Link prediction in drug-target interaction network using similarity indices[J].BMC Bioinformatics,2017,18(1):39-1-9.
[10] PECH R,HAO D,PAN L M,et al.Link prediction via matrix completion[J].EPL,2017,117(3):38002-1-9.
[11] ADITYA K M,CHARLES E.Link prediction via matrix factorization[C]//ECML PKDD’11 Proceedings of the 2011 European Conference on Machine Learning and Knowledge Discovery in Databases-Volume Part Ⅱ.Berlin:Spring-Verlag,2011:437-453.
[12] 吴世伟,熊赟.基于概率矩阵分解的个性化链接预测算法[J].计算机应用与软件,2015,32(8):243-247.
[13] ZHAI S F,ZHANG Z F.Dropout training of matrix factorization and auto encoder for link prediction in sparse graphs[EB/OL].[2015-12-14].http://arxiv.org/abs/1512.04483.
[14] BRYAN P,RAMI A,STEVEN S.Deepwalk:online learning of social representations[EB/OL].[2014-03-26].http://arxiv.org/abs/1403.6652.
[15] HE X N,LIAO L Z,ZHANG H W,et al.Neural collaborative filtering[C]//Proceedings of the 26th Interactional Conference on World Wide Web.[S.l.:s.n.],2017:173-182.
[16] STEFFEN R,CHRISTOPH F,ZENO G,et al.BPR:bayesian personalized ranking from implicit feedback[EB/OL].[2012-05-09].http://arxiv.org/abs/1205.2618.
[17] ARTUS K,LUCAS D,CHRISTOPH F,et al.Multi-relational matrix factorization using bayesian personalized ranking for social network data[C]//Proceedings of the fifth ACM interactional conference on Web search and data mining.New York:ACM,2012:173-182.
[18] RUCK D W,ROGERS S K,KABRISKY M,et al.The multilayer perceptron as an approximation to a Bayes optimal discriminant function[J].IEEE Transactions on Neural Networks,1990,1(4):296-298.
[19] JAMALI M,ESTER M.A matrix factorization technique with trust propagation for recommendation in social networks[C]//Proceeding of the fourth ACM conference on Recommender systems.New York,ACM,2010:135-142.
[20] MA H,ZHOU D Y,LIU C,et al.Recommender systems with social regularization[C]//Proceedings of the Fourth ACM International Conference on Web Search and Data Mining.Nwe York:ACM,2011:287-296.
[21] BECK J E,WOOLF B P.High-level student modeling with machine learning[C]//International Conference on Intelligent Tutoring Systems.Berlin Heidelberg:Springer,2000:584-593.
[22] HAHNLOSER R,SPARPESHKAR R,MAHOWALD M A,et al.Digital selection and analogue amplification coexist in a cortexinspired silicon circuit[J].Nature,2000,405(6789):947-951.
[23] ADAMIC L A,ADAR E.Friends and neighbors on the Web[J].Social networks,2003,25(3):211-230.
[24] HE X N,CHEN T,KAN M Y,et al.Tri rank:review-aware explainable recommendation by modeling aspects[C]//CIKM2015 Proceedings of the 24th ACM International on Conference on Information and Knowledge Management.New York:ACM,2015:1661-1670.
[25] SRIVASTAVA N,HINTON G,KRIZHVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.
[2] 张宗宇.社会化网络的链接预测[D].北京:北京邮电大学,2011.
[3] O′MADADHAIN J,HUTCHINS J,SMYTH P.Prediction and ranking algorithms for event-based network data[M].New York:ACM Press,2005:23-30.
[4] LU L Y,JIN C H,ZHOU T.Similarity index based on local paths for link prediction of complex networks[J].Physical Review E,2009,80(1):046122.
[5] LIU W P,LU L Y.Link prediction based on local random walk[J].A letters Journal Exploring the Frontiers of Physics,2010,89(5):58007-58012.
[6] 孙诚,王志海.社会网络中基于神经网络的链路预测方法[J].数学建模及其应用,2017,4(6):10-17.
[7] FEI T,XIA Y X,ZHU B Y.Link prediction in complex networks:a mutual information Perspective[J].PloS ONE,2014,9(9):e107056.
[8] ZHU B Y,XIA Y X.Link prediction in weighted networks:a weighted mutmal information model[J].PLoS ONE,2016,11(2):e0148265.
[9] LU Y D,GUO Y F,KNORHONEN A.Link prediction in drug-target interaction network using similarity indices[J].BMC Bioinformatics,2017,18(1):39-1-9.
[10] PECH R,HAO D,PAN L M,et al.Link prediction via matrix completion[J].EPL,2017,117(3):38002-1-9.
[11] ADITYA K M,CHARLES E.Link prediction via matrix factorization[C]//ECML PKDD’11 Proceedings of the 2011 European Conference on Machine Learning and Knowledge Discovery in Databases-Volume Part Ⅱ.Berlin:Spring-Verlag,2011:437-453.
[12] 吴世伟,熊赟.基于概率矩阵分解的个性化链接预测算法[J].计算机应用与软件,2015,32(8):243-247.
[13] ZHAI S F,ZHANG Z F.Dropout training of matrix factorization and auto encoder for link prediction in sparse graphs[EB/OL].[2015-12-14].http://arxiv.org/abs/1512.04483.
[14] BRYAN P,RAMI A,STEVEN S.Deepwalk:online learning of social representations[EB/OL].[2014-03-26].http://arxiv.org/abs/1403.6652.
[15] HE X N,LIAO L Z,ZHANG H W,et al.Neural collaborative filtering[C]//Proceedings of the 26th Interactional Conference on World Wide Web.[S.l.:s.n.],2017:173-182.
[16] STEFFEN R,CHRISTOPH F,ZENO G,et al.BPR:bayesian personalized ranking from implicit feedback[EB/OL].[2012-05-09].http://arxiv.org/abs/1205.2618.
[17] ARTUS K,LUCAS D,CHRISTOPH F,et al.Multi-relational matrix factorization using bayesian personalized ranking for social network data[C]//Proceedings of the fifth ACM interactional conference on Web search and data mining.New York:ACM,2012:173-182.
[18] RUCK D W,ROGERS S K,KABRISKY M,et al.The multilayer perceptron as an approximation to a Bayes optimal discriminant function[J].IEEE Transactions on Neural Networks,1990,1(4):296-298.
[19] JAMALI M,ESTER M.A matrix factorization technique with trust propagation for recommendation in social networks[C]//Proceeding of the fourth ACM conference on Recommender systems.New York,ACM,2010:135-142.
[20] MA H,ZHOU D Y,LIU C,et al.Recommender systems with social regularization[C]//Proceedings of the Fourth ACM International Conference on Web Search and Data Mining.Nwe York:ACM,2011:287-296.
[21] BECK J E,WOOLF B P.High-level student modeling with machine learning[C]//International Conference on Intelligent Tutoring Systems.Berlin Heidelberg:Springer,2000:584-593.
[22] HAHNLOSER R,SPARPESHKAR R,MAHOWALD M A,et al.Digital selection and analogue amplification coexist in a cortexinspired silicon circuit[J].Nature,2000,405(6789):947-951.
[23] ADAMIC L A,ADAR E.Friends and neighbors on the Web[J].Social networks,2003,25(3):211-230.
[24] HE X N,CHEN T,KAN M Y,et al.Tri rank:review-aware explainable recommendation by modeling aspects[C]//CIKM2015 Proceedings of the 24th ACM International on Conference on Information and Knowledge Management.New York:ACM,2015:1661-1670.
[25] SRIVASTAVA N,HINTON G,KRIZHVSKY A,et al.Dropout:a simple way to prevent neural networks from overfitting[J].The Journal of Machine Learning Research,2014,15(1):1929-1958.