-
摘要:
深入分析电商行业的用户个性化数据并提供推荐服务近年来已成为业界的热点。推荐服务的基础是对用户的潜在兴趣进行挖掘，并对商品的感兴趣程度进行预测。因此，以此为背景，研究用户对商品的评分预测。对电商业的关系型数据在推荐系统中的应用进行了研究，提出了通过使用网络表示学习进行评分预测的方法。首先，将关系型数据构建成异构网络，用户和商品为网络中的节点。然后，设计了兼顾网络结构信息和节点之间相似性的个性化异构网络采样方法，并对节点进行表示学习。最后，将学习到的用户、商品表示向量输入到神经网络中进行训练，利用优化后的神经网络模型进行评分预测。实验结果表明：所提方法在YELP 13、Movielens 100k、Movielens 1m数据集上都有较高的准确率，对比常用方法，准确率提升6.5%以上。
Abstract:In recent years, it has become a hot spot to deeply analyze the personalized data of e-commerce users and provide recommendation services.The basis of recommendation service is to mine the potential interest of users and predict user's interest of products. Therefore, this paper takes this as the background to study the user's rating prediction of products. This paper studies the application of relational data of e-commerce in recommendation system, and puts forward a method of rating prediction by using network representation learning. First, the relational data is constructed into a heterogeneous network, and the users and items are the nodes in the network. Then, a personalized heterogeneous network sampling method is designed, which takes into account the network structure information and the similarity between nodes, and the nodes are represented and learned. Finally, the learned user and items representation vectors are input into the neural network for training, and the optimized neural network model is used to predict the score. The experimental results show that this method has high accuracy on YELP 13, Movielens 100k and Movielens 1m datasets. Compared with common methods, the accuracy is improved by more than 6.5%.
-
Key words:
- rating prediction /
- recommendation system /
- representation learning /
- random walk /
- neural network
-
图 4 Movielens数据集中用户分布
Figure 4. Schematic diagram of user distribution in Movielens dataset
图 7 不同随机游走序列长度对实验结果的影响
Figure 7. Effect of random walk sequence lengths on experimental results
表 1 数据集基本信息
Table 1. Basic information of datasets
数据集 用户
个数商品
个数评论
条数用户平均
评论数商品平均
被评论数Movielens 100k 943 1 682 100 000 106.04 59.45 Movielens 1m 6 040 3 900 1 000 209 165.59 256.46 YELP 13 18 417 9 130 103 090 5.59 11.29 
下载: 导出CSV
表 2 不同数据集上的实验结果
Table 2. Experimental results on different datasets
数据集 ACC0 ACC1 ACC2 MAE Movielens 1m 0.419 5 0.872 5 0.978 2 0.735 Movielens 100k 0.419 3 0.864 2 0.975 2 0.743 YELP 13 0.433 3 0.829 1 0.933 3 0.624
下载: 导出CSV
表 3 Movielens 100k数据集上的实验结果
Table 3. Experimental results on Movielens 100k dataset
方法 MAE 基于矩阵分解 0.860 基于用户属性和评分 0.752 MFCF 0.750 RLES 0.821 RLCOS 0.830 RLPCC 0.822 RLNN 0.743
下载: 导出CSV
-
[1] LI G, ZHANG Z, WANG L, et al.One-class collaborative filtering based on rating prediction and ranking prediction[J].Knowledge-Based Systems, 2017, 124:46-54. doi: 10.1016/j.knosys.2017.02.034 [2] MARGARIS D, VASILOPOULOS D, VASSILAKIS C, et al.Improving collaborative filtering's rating prediction coverage in sparse datasets through the introduction of virtual near neighbors[C]//201910th International Conference on Information, Intelligence, Systems and Applications(IISA).Piscataway: IEEE Press, 2019: 1-8. [3] LIU H, HU Z, MIAN A, et al.A new user similarity model to improve the accuracy of collaborative filtering[J].Knowledge-Based Systems, 2014, 56:156-166. doi: 10.1016/j.knosys.2013.11.006 [4] GANU G, ELHADAD N, MARIAN A.Beyond the stars: Improving rating predictions using review text content[C]//International Workshop on the Web and Databases, 2009: 1-6. [5] FIKIR O B, YAZ Ï O, ÖZYER T.Movie rating prediction with matrix factorization algorithm[M].Berlin:Springer, 2013:631-643. [6] 杨阳, 向阳, 熊磊.基于矩阵分解与用户近邻模型的协同过滤推荐算法[J].计算机应用, 2012, 32(2):395-398. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201202025.htmYANG Y, XIANG Y, XIONG L.Collaborative filtering and recommendation algorithm based on matrix factorization and user nearest neighbor model[J].Journal of Computer Applications, 2012, 32(2):395-398(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201202025.htm [7] CHAMBUA J, NIU Z, YOUSIF A, et al.Tensor factorization method based on review text semantic similarity for rating prediction[J].Expert Systems with Applications, 2018, 114:629-638. doi: 10.1016/j.eswa.2018.07.059 [8] KOREN Y, BELL R, VOLINSKY C.Matrix factorization techniques for recommender systems[J].Computer, 2009, 42(8):30-37. doi: 10.1109/MC.2009.263 [9] VIARD T, FOURNIER-S'NIEHOTTA R.Movie rating prediction using content-based and link stream features[EB/OL].(2018-05-08)[2020-03-01].https://arxiv.org/abs/1805.02893v1. [10] QIAO Z, ZHANG P, HE J, et al.Combining geographical information of users and content of items for accurate rating prediction[C]//Proceedings of the Companion Publication of the 23rd International Conference on World Wide Web Companion.New York: ACM Press, 2014: 361-362. [11] 丁少衡, 姬东鸿, 王路路.基于用户属性和评分的协同过滤推荐算法[J].计算机工程与设计, 2015, 36(2):487-497. https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201502039.htmDING S H, JI D H, WANG L L.Collaborative filtering recommendation algorithm based on user attributes and scores[J].Computer Engineering and Design, 2015, 36(2):487-497(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-SJSJ201502039.htm [12] 邓日升, 岳昆, 武浩, 等.面向商品评分预测的隐变量模型构建与推理[J].小型微型计算系统, 2017, 38(2):352-356. https://www.cnki.com.cn/Article/CJFDTOTAL-XXWX201702031.htmDENG R S, YUE K, WU H, et al.Constructing and inferring latent variable model for predicting product ratings[J].Journal of Chinese Computer Systems, 2017, 38(2):352-356(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-XXWX201702031.htm [13] DAVOUDI A, CHATTERJEE M.Product rating prediction using trust relationships in social networks[C]//201613th IEEE Annual Consumer Communications & Networking Conference (CCNC).Piscataway: IEEE Press, 2016: 15887960. [14] QUIJANO-SÁNCHEZ L, RECIO-GARCÍA J A, DÍAZ-AGUDO B.Group recommendation methods for social network environments[C]//3rd Workshop on Recommender Systems and the Social Web within the 5th ACM International Conference on Recommender Systems.New York: ACM Press, 2011: 24. [15] 肖志宇, 翟玉庆.改进的基于信任网络和随机游走策略的评分预测模型[J].南京理工大学学报, 2015, 39(5):602-608. https://www.cnki.com.cn/Article/CJFDTOTAL-NJLG201505015.htmXIAO Z Y, ZHAI Y Q.Improved rating prediction model basing on trust network and random walk strategy[J].Journal of Nanjing University of Science and Technology, 2015, 39(5):602-608(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-NJLG201505015.htm [16] DAVOUDI A, CHATTERJEE M.Social trust model for rating prediction in recommender systems:Effects of similarity, centrality, and social ties[J].Online Social Networks and Media, 2018, 7:1-11. doi: 10.1016/j.osnem.2018.05.001 [17] 周文乐, 朱明, 蒋旦.综合时间及评分因素的电影评分预测方法[J].电子技术, 2015, 44(8):72-77. doi: 10.3969/j.issn.1000-0755.2015.08.021ZHOU W L, ZHU M, JIANG D.Time and rating factor considered rating prediction method[J].Electronic Technology, 2015, 44(8):72-77(in Chinese). doi: 10.3969/j.issn.1000-0755.2015.08.021 [18] TIROSHI A, BERKOVSKY S, KAAFAR M A, et al.Improving business rating predictions using graph based features[C]//Proceedings of the 19th International Conference on Intelligent User Interfaces.New York: ACM Press, 2014: 17-26. [19] PEROZZI B, AL-RFOU R, SKIENA S.DeepWalk: Online learning of social representations[C]//Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.New York: ACM Press, 2014: 701-710. [20] MIKOLOV T, CHEN K, CORRADO G, et al.Efficient estimation of word representations in vector space[EB/OL].(2013-01-16)[2020-03-01].https://arxiv.org/abs/1301.3781. [21] ORTEGA F, HERNANDO A, BOBADILLA J, et al.Recommending items to group of users using matrix factorization based collaborative filtering[J].Information Sciences, 2016, 345:313-324. doi: 10.1016/j.ins.2016.01.083 -
下载:
点击查看大图
计量
- 文章访问数: 621
- HTML全文浏览量: 135
- PDF下载量: 83
- 被引次数: 0
