One Class Collaborative Filtering Recommendation Algorithm Coupled with User Common Characteristics

doi:10.3778/j.issn.1673-9418.2009011

Abstract

Abstract:

Combining explicit features with implicit feedback is a common method to improve the recommendation accuracy of one class collaborative filtering (OCCF). However, current studies generally integrate the original explicit features or cross features directly into OCCF models, which makes it difficult to determine which explicit features are really vital, so it is untoward to achieve significant performance improvement. To sum up, a one class collaborative filtering recommendation algorithm coupled with user common characteristics (UCC-OCCF) is proposed. First, the neighbor-based common preference representation network (NB-CPR) is established to learn the interaction between users with similar explicit characteristics as the current users and a certain type of item, and to indirectly use explicit characteristics to obtain common preferences. Then, the deep latent factors representation (DLFR) uses a deep neural network to learn the potential factors between the user and the item, thus obtaining the interaction probability between the current user and the item. At last, the NB-CPR is combined with the personal depth latent factor representation network for training, so as to couple the common characteristics of users into OCCF model to improve the recommendation accuracy. Experimental results on public datasets MovieLens 100K, MovieLens 1M and MyAnimelist, show that UCC-OCCF can significantly improve the recommendation accuracy of OCCF.

Key words: one-class collaborative filtering (OCCF), deep learning, common preferences, implicit feedback, explicit feature

摘要：

将显式特征与隐式反馈相结合是提高单类协同过滤（OCCF）推荐准确性的常用方法。但目前的研究一般是直接将原始显式特征或交叉特征集成到OCCF模型中,因其难以判断哪些显式特征是真正重要的,故很难获得显著的性能改进。基于此,提出了一种耦合用户公共特征的单类协同过滤推荐算法（UCC-OCCF）。首先,建立基于邻居的共同偏好表示网络（NB-CPR）,学习与当前用户具有相似显式特征的邻居用户和某一类项目之间的交互关系,间接利用显式特征以获得共同偏好;然后,建立个人深度潜在因素表示网络（DLFR）,使用深度神经网络学习用户-项目之间的潜在因素,从而得到当前用户与项目之间的交互概率;最后,基于邻居的共同偏好表示网络与个人深度潜在因素表示网络进行联合训练,从而将用户公共特征耦合到单类协同过滤推荐模型中,以提高推荐准确度。在公共数据集MovieLens 100K、MovieLens 1M和MyAnimelist上的实验结果表明,UCC-OCCF可以显著提高OCCF的推荐准确性。

关键词: 单类协同过滤（OCCF）, 深度学习, 共同偏好, 隐式反馈, 显式特征

CLC Number:

TP18

ZHANG Quangui, HU Jiayan, WANG Li. One Class Collaborative Filtering Recommendation Algorithm Coupled with User Common Characteristics[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(3): 637-648.

张全贵, 胡嘉燕, 王丽. 耦合用户公共特征的单类协同过滤推荐算法[J]. 计算机科学与探索, 2022, 16(3): 637-648.

Figures/Tables 17

Fig.1 User-item interaction matrix with implicit (one-class) feedback

Table 1 Mathematical notations

符号	说明
$U$	用户集
$V$	项目集
$u$	用户集 $U$ 的一个实例
$v$	项目集 $V$ 的一个实例
$p$	用户 $u$ 的潜在向量
$q$	项目 $v$ 的潜在向量
$p i$	向量 $p$ 的第 $i$ 个元素
$q j$	向量 $q$ 的第 $j$ 个元素
$W$	全连接层的权值矩阵
$W N B i$	第 $i$ 个邻居的权重
$N B i$	当前用户 $u$ 的第 $i$ 个邻居
$y (i)$	训练/测试数据集的第 $i$ 个样本的标签
$y ⌢ (i)$	测试数据集的第 $i$ 个样本的预测值

Table 1 Mathematical notations

符号	说明
$U$	用户集
$V$	项目集
$u$	用户集 $U$ 的一个实例
$v$	项目集 $V$ 的一个实例
$p$	用户 $u$ 的潜在向量
$q$	项目 $v$ 的潜在向量
$p i$	向量 $p$ 的第 $i$ 个元素
$q j$	向量 $q$ 的第 $j$ 个元素
$W$	全连接层的权值矩阵
$W N B i$	第 $i$ 个邻居的权重
$N B i$	当前用户 $u$ 的第 $i$ 个邻居
$y (i)$	训练/测试数据集的第 $i$ 个样本的标签
$y ⌢ (i)$	测试数据集的第 $i$ 个样本的预测值

Fig.2 UCC-OCCF framework

Fig.3 Deep latent factor representation (DLFR）

Fig.4 User similarity matrix and neighbors

Table 2 Statistics of datasets

数据集	用户	项目	交互
MovieLens 100K	943	1 682	100 000
MovieLens 1M	6 040	3 592	1 000 209
MyAnimeList	9 130	14 478	2 945 242

Table 3 Explicit features of users and items

类别	可见的特性
用户	性别、年龄、职业、邮编
项目	标题类型

Table 4 Test hyper-parameters

测试参数	测试参数具体值
嵌入层的维度	{8,16,32,64,128,256}
最后隐藏层维度	{8,16,32,64,128,256}
学习率	{0.000 1,0.000 5,0.001 0,0.003 0,0.005 0,0.010 0}
批处理大小	{128,256,512,1 024}
用户邻居数量	{10,20,30,40,50,100}

Fig.5 Evaluation of HR@K

Fig.6 Evaluation of NDCG@K

Fig.7 Evaluation of MRR

Fig.8 Evaluation of HR@10 using different number of neighbors

Fig.9 Evaluation of NDCG@10 using different number of neighbors

Fig.10 Evaluation of MRR using different number of neighbors

Fig.11 Evaluation of model structure

Table 5 Time for training and prediction

数据集	时长/s	itemKNN	Wide&Deep	NeuMF	NGCF	DeepICF	UCC-OCCF-COS	UCC-OCCF-HAM	UCC-OCCF-Pear
MovieLens 100K	训练（epoch）	12.3	24.5	13.1	14.4	20.3	25.7	20.6	28.6
	训练总时长	293.4	575.8	510.6	432.6	512.4	421.3	389.6	458.6
	预测总时长	1.5	2.0	1.6	3.9	2.0	2.1	1.9	2.3
MovieLens 1M	训练（epoch）	125.1	325.1	208.1	805.5	308.6	425.3	456.2	430.0
	训练总时长	3 481.7	6 502.5	4 994.4	19 332.2	6 058.6	4 379.5	4 453.6	4 620.3
	预测总时长	11.6	18.6	11.3	21.6	13.0	15.6	20.0	17.3
MyAnimeList	训练（epoch）	389.4	653.4	442.9	1 970.1	555.2	582.4	568.4	545.9
	训练总时长	10 903.2	11 454.4	8 035.1	43 342.2	12 510.6	3 510.8	3 814.4	2 183.6
	预测总时长	12.5	12.5	13.9	37.9	20.1	18.8	17.8	15.3

Table 6 Hyper-parameters

超参数	MovieLens 100K	MovieLens 1M	MyAnimeList
嵌入层的维度	8	64	64
最后隐藏层维度	8	256	256
学习率	0.000 5	0.001 0	0.000 1
批处理大小	256	256	512

References 33

[1]	KOREN Y, BELL R M, VOLINSKY C. Matrix factorization techniques for recommender systems[J]. Computer, 2009, 42(8): 30-37.
[2]	PAN W K, CHEN L. Group Bayesian personalized ranking with rich interactions for one-class collaborative filtering[J]. Neurocomputing, 2016, 207: 501-510. DOI URL
[3]	RENDLE S, FREUDENTHALER C, GANTNER Z, et al. BPR: Bayesian personalized ranking from implicit feedback[J]. arXiv:1205.2618, 2012.
[4]	ZHOU W, LI J, ZHOU Y, et al. Bayesian pairwise learning to rank via one-class collaborative filtering[J]. Neurocomputing, 2019, 367: 176-187. DOI URL
[5]	俞春花, 刘学军, 李斌. 隐式反馈场景中融合社交信息的上下文感知推荐[J]. 计算机科学, 2016, 43(6): 248-253.
	YU C H, LIU X J, LI B. Implicit feedback personalized recommendation model fusing context-aware and social network process[J]. Computer Science, 2016, 43(6): 248-253.
[6]	YAO W L, HE J, HUANG G Y, et al. A graph-based model for context-aware recommendation using implicit feedback data[J]. World Wide Web, 2015, 18(5): 1351-1371. DOI URL
[7]	PAN R, ZHOU Y H, CAO B, et al. One-class collaborative filtering[C]// Proceedings of the 8th IEEE International Conference on Data Mining, Pisa, Dec 15-19, 2008. Washington: IEEE Computer Society, 2008: 502-511.
[8]	PAN W K, YANG Q, CAI W L, et al. Transfer to rank for heterogeneous one-class collaborative filtering[J]. ACM Transactions on Information Systems, 2019, 37(1): 1-20.
[9]	JAVED F, HAYAT M. Predicting subcellular localization of multi-label proteins by incorporating the sequence features into Chou’s PseAAC[J]. Genomics, 2019, 111(6): 1325-1332. DOI URL
[10]	CHENG H T, KOC L, HARMSEN J, et al. Wide & deep learning for recommender systems[C]// Proceedings of the 1st Workshop on Deep Learning for Recommender Systems, Boston, Sep 15, 2016. New York: ACM, 2016: 7-10.
[11]	GUO H, TANG R, YE Y, et al. DeepFM: a factorization-machine based neural network for CTR prediction[J]. arXiv:1703.04247, 2017.
[12]	BAG S, KUMAR S K, TIWARI M K. An efficient recommendation generation using relevant Jaccard similarity[J]. Information Sciences, 2019, 483: 53-64. DOI URL
[13]	HE X N, LIAO L Z, ZHANG H W, et al. Neural collaborative filtering[C]// Proceedings of the 26th International Conference on World Wide Web, Perth, Apr 3-7, 2017. New York: ACM, 2017: 173-182.
[14]	WANG N Y, YEUNG D Y. Learning a deep compact image representation for visual tracking[C]// Proceedings of the 27th Annual Conference on Neural Information Processing Systems, Lake Tahoe, Dec 5-8, 2013. Red Hook: Curran Associates, 2013: 809-817.
[15]	KALCHBRENNER N, GREFENSTETTE E, BLUNSOM P. A convolutional neural network for modelling sentences[J]. arXiv:1404. 2188, 2014.
[16]	VAN DEN OORD A, DIELEMAN S, SCHRAUWEN B. Deep content-based music recommendation[C]// Proceedings of the 27th Annual Conference on Neural Information Processing Systems, Lake Tahoe, Dec 5-8, 2013. Red Hook: Curran Associates, 2013: 2643-2651.
[17]	MIRBAKHSH N, LING C X. Leveraging clustering to improve collaborative filtering[J]. Information Systems Frontiers, 2018, 20(1): 111-124. DOI URL
[18]	SIDANA S, TROFIMOV M, HORODNITSKII O, et al. Representation learning and pairwise ranking for implicit feedback in recommendation systems[J]. arXiv:1705.00105, 2017.
[19]	ELKAHKY A M, SONG Y, HE X D. A multi-view deep learning approach for cross domain user modeling in recommendation systems[C]// Proceedings of the 24th International Conference on World Wide Web, Florence, May 18-22, 2015. New York: ACM, 2015: 278-288.
[20]	SEDHAIN S, BUI H H, KAWALE J, et al. Practical linear models for large-scale one-class collaborative filtering[C]// Proceedings of the 25th International Joint Conference on Artificial Intelligence, New York, Jul 9-15, 2016. Menlo Park: AAAI, 2016: 3854-3860.
[21]	MIKOLOV T, CHEN K, CORRADO G, et al. Efficient estimation of word representations in vector space[C]// Proceedings of the 1st International Conference on Learning Representations, Scottsdale, May 2-4, 2013: 1301-3781.
[22]	MIKOLOV T, SUTSKEVER I, CHEN K, et al. Distributed representations of words and phrases and their compositionality[C]// Proceedings of the 27th Annual Conference on Neural Information Processing Systems, Lake Tahoe, Dec 5-8, 2013. Red Hook: Curran Associates, 2013: 3111-3119.
[23]	WANG C, DONG X J, ZHOU F, et al. Coupled attribute similarity learning on categorical data[J]. IEEE Transactions on Neural Networks and Learning Systems, 2015, 26(4): 781-797. DOI URL
[24]	LIU M S, PAN W K, LIU M, et al. Mixed similarity learning for recommendation with implicit feedback[J]. Knowledge Based Systems, 2017, 119: 178-185. DOI URL
[25]	HARPER F M, KONSTAN J A. The MovieLens datasets: history and context[J]. ACM Transactions on Interactive Intelligent Systems, 2016, 5(4): 1-19.
[26]	SARWAR B M, KARYPIS G, KONSTAN J A, et al. Item-based collaborative filtering recommendation algorithms[C]// Proceedings of the 10th International World Wide Web Conference, Hong Kong, China, May 1-5, 2001. New York: ACM, 2001: 285-295.
[27]	WANG X, HE X N, WANG M, et al. Neural graph collaborative filtering[C]// Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Paris, Jul 21-25, 2019. New York: ACM, 2019: 165-174.
[28]	XUE F, HE X N, WANG X, et al. Deep item-based collaborative filtering for top-N recommendation[J]. ACM Transactions on Information Systems, 2019, 37(3): 1-25.
[29]	HE X N, ZHANG H W, KAN M Y, et al. Fast matrix factorization for online recommendation with implicit feedback[C]// Proceedings of the 39th International ACM SIGIR Conference on Research and Development in Information Retrieval, Pisa, Jul 17-21, 2016. New York: ACM, 2016: 549-558.
[30]	BAYER I, HE X N, KANAGAL B, et al. A generic coordinate descent framework for learning from implicit feedback[C]// Proceedings of the 26th International Conference on World Wide Web, Perth, Apr 3-7, 2017. New York: ACM, 2017: 1341-1350.
[31]	KOREN Y. Factorization meets the neighborhood: a multifaceted collaborative filtering model[C]// Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Las Vegas, Aug 24-27, 2008. New York: ACM, 2008: 426-434.
[32]	HE X N, CHEN T, KAN M Y, et al. TriRank: review-aware explainable recommendation by modeling aspects[C]// Proceedings of the 24th ACM International Conference on Information and Knowledge Management, Melbourne, Oct 19-23, 2015. New York: ACM, 2015: 1661-1670.
[33]	SAKIB N, AHMAD R B, HARUNA K. A collaborative approach toward scientific paper recommendation using citation context[J]. IEEE Access, 2020, 8: 51246-51255. DOI URL