从用户行为数据构建用户偏好模型，是解决个性化服务、评分预测和用户行为定向等问题的重要基础.本文从用户的评分数据出发，以多个隐变量分别描述用户在评分对象多个维度的偏好，以含有多个隐变量的贝叶斯网（简称隐变量模型）作为表示用户偏好的基本知识框架.首先根据用户偏好和隐变量的特定含义给出模型构建的约束条件，进而提出基于约束条件的模型构建方法，使用约束条件下的EM算法来计算模型参数，约束条件下的SEM算法来构建模型结构.针对多隐变量情形下模型构建过程中产生大量中间数据带来的计算复杂度急剧上升的问题，本文使用Spark计算框架实现模型构建的方法.建立在Movielens数据集上的实验表明，本文提出的方法是有效的.
Modeling user preference from user behavior data is the basis of personalization service, score prediction, user behavior targeting, etc. In this paper, multi-dimensional preferences from rating data are described by multiple latent variables and the Bayesian network with multiple latent variables is adopted as the preliminary knowledge framework of user preference. Constraint conditions are given according to the inherence of user preference and latent variables, upon which we propose a method for modeling user preference. Parameters are computed by EM algorithm and structure is established by SEM algorithm with respect to the given constraints. In the case of multiple latent variables, a large amount of intermediate data is generated in modeling, which causes the increasing computational complexity. Therefore, we implement the modeling method with Spark computing framework. Experiments results on the Movielens dataset verify that the method proposed in this paper is effective.
[1] GROUPLENS RESEARCH.MovieLensDataset[EB/OL].[2017-08-20]. http://grouplens.org/datasets/movielens/.
[2] KASSAK O, KOMPAN M, BIELIKOVA M. User preference modeling by global and individual weights for personalized recommendation[J]. Acta Polytechnica Hungarica, 2015, 12(8):27-41
[3] YIN H, CUI B, CHEN L, et al. Modeling location-based user rating profiles for personalized recommendation[J]. ACM Transactions on Knowledge Discovery from Data, 2015, 9(3):1-41.
[4] YUAN Q, CONG G, MA Z, et al. Who, where, when and what:Discover spatio-temporal topics for Twitter users[C]//ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2013:605-613.
[5] ZHAO K, CONG G, YUAN Q, et al. SAR:A sentiment-aspect-region model for user preference analysis in geo-tagged reviews[C]//IEEE, International Conference on Data Engineering. IEEE, 2015:675-686.
[6] JIANG B, LIANG J, SHA Y, et al. Retweetingbehavior prediction based on one-class collaborative filtering in social networks[C]//Proceedings of the 39th International ACM Special Interest Group on Information Retrievalconference on Research and Development in Information Retrieval. ACM, 2016:977-980.
[7] 张连文, 郭海鹏. 贝叶斯网引论[M]. 北京:科学出版社, 2006.
[8] DAPHNEKOLLER, NIRFRIEDMAN,
[9] KIM J S, JUN C H. Ranking evaluation of institutions based on a Bayesian network having a latent variable[J]. Knowledge-Based Systems, 2013, 50:87-99.
[10] SCHÜTZ W, SCHÄFER R. Bayesian networks for estimating the user's interests in the context of a configuration task[C]//Proceedings of the UM2001 Workshop on Machine Learning for User Modeling, 2001:13-17.
[11] FRIEDMAN N. The Bayesian structural EM algorithm[C]//Proceedings of the 14th Conference on Uncertainty in Artificial Intelligence, 2013:129-138.
[12] ELIDAN G, FRIEDMAN N. Learning hidden variable networks:The information bottleneck approach[J]. Journal of Machine Learning Research, 2005, 6(6):81-127.
[13] JIN C, ZHANG Y, BALAKRISHNAN S, et al. Local maxima in the likelihood of gaussian mixture models:Structural results and algorithmic consequences[J]. Advances in Neural Information Processing Systems, 2016, 4(1):16-24.
[14] ZHAO G, QIANX, XIE X. User-service rating prediction by exploring social users' rating behaviors[J]. IEEE Transactions on Multimedia, 2016, 18(3):496-506.
[15] 高全力, 高岭, 杨建锋, 等. 上下文感知推荐系统中基于用户认知行为的偏好获取方法[J]. 计算机学报, 2015(9):1767-1776.
[16] 王红兵, 孙文龙, 王华兰. Web服务选择中偏好不确定问题的研究[J]. 计算机学报, 2013, 36(2):275-285.
[17] 史艳翠, 孟祥武, 张玉洁, 等. 一种上下文移动用户偏好自适应学习方法[J]. 软件学报, 2012, 23(10):2533-2549.
[18] GAO R, YUE K, WU H, et al. Modeling user preference from rating data based on the bayesian network with a latent variable[C]//Proceedings of 17th International Conference on Web-Age Information Management, 2016:3-16.
[19] HUANG Y, BIAN L. A Bayesian network and analytic hierarchy process based personalized recommendations for tourist attractions over the Internet[J]. Expert Systems with Applications, 2009, 36(1):933-943.
[20] CHAPELLE O, ZHANG Y. A dynamic Bayesian network click model for web search ranking[C]//Proceedings of the 18th International Conference on World Wide Web, ACM, 2009:1-10.
[21] HUETE J, DE CAMPOS L M, FERNANDEZ-LUNA J M, et al. Using structural content information for learning user profiles[C]//Proceedings of 30th Special Interest Group on Information Retrieval, 2007:38-45.
[22] AUFFENBERG F, STEIN S, ROGERS A. A personalised thermal comfort model using a Bayesian network[C]//Proceedings of the 2015 International Joint Conference on Artificial Intelligence, 2015:130-139.
[23] YUE K, FANG Q, WANG X, et al. A parallel and incremental approach for data-intensive learning of Bayesian networks[J]. IEEE Transactions on Cybernetics, 2015, 45(12):2890-2909.
[24] TAMADA Y, IMOTO S, MIYANO S. Parallel algorithm for learning optimal Bayesian network structure[J]. Journal of Machine Learning Research, 2011, 12(7):2437-2459.
[25] MIT. FullBNT[CP/OL].[2017-08-20]. http://www.cs.ubc.ca/murphyk/Software/BNT/FullBNT-1.0.4.Zip.
中国综合性科技类核心期刊(北大核心)