基于用户多特征融合的个性化推荐模型

doi:10.11959/j.issn.1000-0801.2023111

电信科学 ›› 2023, Vol. 39 ›› Issue (5): 101-115.doi: 10.11959/j.issn.1000-0801.2023111

基于用户多特征融合的个性化推荐模型

卢敏¹, 胡娟¹^,², 张先超², 丁伟健¹^,², 乐光学²

¹ 江西理工大学理学院，江西赣州 341000
² 浙江省医学电子与数字健康重点实验室，浙江嘉兴 314001

修回日期:2023-05-11 出版日期:2023-05-20 发布日期:2023-05-01
作者简介:卢敏（1964- ），男，江西理工大学教授，主要研究方向为深度学习和智能通信、电子材料和设备
胡娟（1997- ），女，江西理工大学硕士生，主要研究方向为深度学习和智能通信
张先超（1984- ），男，博士，浙江省医学电子与数字健康重点实验室研究员，主要研究方向为无线网络资源管理、人工智能等
丁伟健（1993- ），男，江西理工大学硕士生，主要研究方向为深度学习和智能通信
乐光学（1963- ），男，博士，嘉兴学院信息科学与工程学院、浙江省医学电子与数字健康重点实验室教授，主要研究方向为多云融合与协同服务、边缘计算、深度学习、智能通信等
基金资助:
国家自然科学基金资助项目(11704163)

Personalized recommendation model based on users multi-features fusion

Min LU¹, Juan HU¹^,², Xianchao ZHANG², Weijian DING¹^,², Guangxue YUE²

¹ School of Science, Jiangxi University of Science and Technology, Ganzhou 341000, China
² Key Laboratory of Medical Electronics and Digital Health of Zhejiang Province, Jiaxing 314001, China

Revised:2023-05-11 Online:2023-05-20 Published:2023-05-01
Supported by:
The National Natural Science Foundation of China(11704163)

摘要/Abstract

摘要：

个性化推荐是提取特定信息的有效手段之一，针对传统推荐法中缺少用户特征问题，提出一种基于多特征融合的广义矩阵分解与深度长短期记忆网络混合推荐模型（hybrid recommendation model of generalized matrix factorization and deep long short-term memory network based on multi-features fusion，HMF）。该模型利用潜在特征向量维度表征年龄、职业对项目的选择性影响，凸显用户个性化；利用长短期记忆（long short-term memory，LSTM）获取用户与项目间的时序特征，再通过多层感知机挖掘其深层次的时序非线性高阶交互关系；将广义矩阵分解（generalized matrix factorization，GMF）获得的简单非线性低阶交互关系与复杂时序非线性高阶交互关系融合，经全连接层得出最终预测评分。HMF有效地利用用户多特征和用户—项目交互信息，实现个性化动态精准推荐。为验证模型的有效性和可行性，在公开数据集MovieLens-1M上进行测试。仿真实验表明，当HMF的潜在特征向量维度为50、MLP层数为7时，HR@10和NDCG@10分别为0.983 1、0.974 9，相比传统单特征模型NCF最优解分别提高了27.61%、54.29%。

关键词: 个性化推荐, 多特征融合, 深度学习

Abstract:

Personalized recommendation is one of the most effective means to extract specific information.Aiming at the problem that lacking of users feature in traditional recommendation methods, a hybrid recommendation model of generalized matrix factorization and deep long short-term memory network based on multi-features fusion (HMF) was proposed.The model used potential eigenvector factors to characterize the selective impact of age and occupation on the project, highlighting user personalization.Long short-term memory (LSTM) was used to obtain the temporal characteristics between users and projects, and then the deep temporal nonlinear higher-order interaction relationship was mined by multi-layer perceptron.The simple nonlinear low-order interaction obtained by generalized matrix factorization (GMF) was fused with the complex time-series nonlinear high-order interaction, and the final prediction score was obtained through the full connection layer.HMF effectively utilized the user’s multi-feature and user-project interaction information to realize personalized dynamic and accurate recommendation.In order to verify the validity and feasibility of the model, the test was conducted on the public dataset MovieLens-1M.The simulation experiment shows that when the potential eigenvector factors of HMF is 50 and the MLP layer is 7, HR@10 and NDCG@10 are 0.983 1 and 0.974 9 respectively, which are 27.61% and 54.29% higher than the optimal solution of the traditional single feature model NCF.

Key words: personalized recommendation, multi-features fusion, deep learning

中图分类号:

TP391.3

卢敏, 胡娟, 张先超, 丁伟健, 乐光学. 基于用户多特征融合的个性化推荐模型[J]. 电信科学, 2023, 39(5): 101-115.

Min LU, Juan HU, Xianchao ZHANG, Weijian DING, Guangxue YUE. Personalized recommendation model based on users multi-features fusion[J]. Telecommunications Science, 2023, 39(5): 101-115.

图/表 12

表1

主要数学标识符号定义"

符号	描述
$U = {u_{1}, u_{2}, \dots, u_{K}}$	用户ID集合
$I = {i_{1}, i_{2}, \dots, i_{W}}$	项目ID集合
$A = {a_{1}, a_{2}, \dots, a_{7}}$	用户年龄集合
$J = {j_{1}, j_{2}, \dots, j_{21}}$	用户职业集合
K	用户ID数目
W	项目ID数目
Y	真实评分
$\tilde{Y}$	预测评分
$m_{u}$	用户ID潜在特征向量
$m_{a}$	用户年龄段潜在特征向量
$m_{j}$	用户职业潜在特征向量
$l_{i}$	项目ID潜在特征向量
$m_{u i} 、 m_{u r}$	LSTM+MLP、GMF模块的用户ID特征向量
$m_{a i} 、 m_{a r}$	LSTM+MLP、GMF模块的用户年龄段特征向量
$m_{j i} 、 m_{j r}$	LSTM+MLP、GMF模块的用户职业特征向量
$l_{i i} 、 l_{i r}$	LSTM+MLP、GMF模块的项目ID特征向量
N	潜在特征向量维度
$U_{K \times N}$	用户ID潜在特征矩阵
$I_{W \times N}$	项目ID潜在特征矩阵
$A_{7 \times N}$	用户年龄段潜在特征矩阵
$J_{21 \times N}$	用户职业潜在特征矩阵

表1

图1

图2

表2

表3

图3

图4

图5

图6

图7

表4

表5

参考文献 29

[1]	LIN Y . A neural network-based approach to personalized recommendation of digital resources[J]. Computer Informatization and Mechanical System, 2022,5(4).
[2]	NAN X , KANATO K , WANG X L . Design and implementation of a personalized tourism recommendation system based on the data mining and collaborative filtering algorithm[J]. Computational Intelligence and Neuroscience, 2022:1424097.
[3]	ZHAN Z , XU B G . Analyzing review sentiments and product images by parallel deep nets for personalized recommendation[J]. Information Processing ＆ Management, 2023,60(1): 103166.
[4]	HUANG Y , GADDE R S K , LOPES S ,et al. Improving occupant thermal comfort through personalized space recommendation[J]. Journal of Computing in Civil Engineering, 2023,37(1): 04022053.
[5]	PANG N . A personalized recommendation algorithm for semantic classification of new book recommendation services for university libraries[J]. Mathematical Problems in Engineering, 2022: 1-8.
[6]	AFOUDI Y , LAZAAR M , ACHHAB M A . Hybrid recommendation system combined content-based filtering and collaborative prediction using artificial neural network[J]. Simulation Modelling Practice and Theory, 2021(113): 102375.
[7]	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. New York:ACM Press, 2017: 173-182.
[8]	DO P M T , NGUYEN T T S . Semantic-enhanced neural collaborative filtering models in recommender systems[J]. Knowledge-Based Systems, 2022(257): 109934.
[9]	RENDLE S . Factorization machines with libFM[J]. ACM Transactions on Intelligent Systems and Technology, 2012,3(3): 1-22.
[10]	SINGH R K , SINGH P K , SINGH J P ,et al. Utilizing alike neighbor influenced similarity metric for efficient prediction in collaborative filter-approach-based recommendation system[J]. Applied Sciences, 2022,12(22): 11686.
[11]	WANG Y , ZHAO X H , ZHANG Z Q ,et al. A collaborative filtering algorithm based on item labels and Hellinger distance for sparse data[J]. Journal of Information Science, 2022,48(6): 749-766.
[12]	DUAN R , JIANG C Q , JAIN H K . Combining review-based collaborative filtering and matrix factorization:a solution to rating’s sparsity problem[J]. Decision Support Systems, 2022(156): 113748.
[13]	SEJWAL V K , ABULAISH M , JAHIRUDDIN . CRecSys:a context-based recommender system using collaborative filtering and LOD[J]. IEEE Access, 2020(8): 158432-158448.
[14]	NGUYEN S T , KWAK H Y , LEE S Y ,et al. Featured hybrid recommendation system using stochastic gradient descent[J]. International Journal of Networked and Distributed Computing, 2021,9(1): 25.
[15]	KIM D , PARK C , OH J ,et al. Convolutional matrix factorization for document context-aware recommendation[C]// Proceedings of the 10th ACM Conference on Recommender Systems. New York:ACM Press, 2016: 233-240.
[16]	田震, 潘腊梅, 尹朴 ,等. 深度矩阵分解推荐算法[J]. 软件学报, 2021,32(12): 3917-3928.
	TIAN Z , PAN L M , YIN P ,et al. Deep matrix factorization recommendation algorithm[J]. Journal of Software, 2021,32(12): 3917-3928.
[17]	王骏, 虞歌 . 基于改进神经协同过滤模型的电影推荐系统[J]. 计算机工程与设计, 2020,41(7): 2069-2075.
	WANG J , YU G . Movie recommendation system based on improved neural collaborative filtering model[J]. Computer Engineering and Design, 2020,41(7): 2069-2075.
[18]	李浩, 张亚钏, 康雁 ,等. 融合循环知识图谱和协同过滤电影推荐算法[J]. 计算机工程与应用, 2020,56(2): 106-114.
	LI H , ZHANG Y C , KANG Y ,et al. Fusion recurrent knowledge graph and collaborative filtering movie recommendation algorithm[J]. Computer Engineering and Applications, 2020,56(2): 106-114.
[19]	高茂庭, 徐彬源 . 基于循环神经网络的推荐算法[J]. 计算机工程, 2019,45(8): 198-202,209.
	GAO M T , XU B Y . Recommendation algorithm based on recurrent neural network[J]. Computer Engineering, 2019,45(8): 198-202,209.
[20]	李利杰, 张君华 . 基于循环神经网络和全局化领域的推荐算法[J]. 计算机与数字工程, 2022,50(8): 1676-1679,1701.
	LI L J , ZHANG J H . Recommendation algorithm based on recurrent neural network and global domain[J]. Computer and Digital Engineering, 2022,50(8): 1676-1679,1701.
[21]	田晓婧, 谢颖华 . 融合长短期记忆网络与广义矩阵分解的神经协同过滤模型[J]. 计算机系统应用, 2022,31(1): 190-194.
	TIAN X J , XIE Y H . Neural collaborative filtering model integrating LSTM and generalized matrix factorization[J]. Computer Systems ＆ Applications, 2022,31(1): 190-194.
[22]	KIRAN R , KUMAR P , BHASKER B . DNNRec:a novel deep learning based hybrid recommender system[J]. Expert Systems With Applications, 2020(144): 113054.
[23]	TANG X L , WANG T Y , YANG H Z ,et al. AKUPM:attention-enhanced knowledge-aware user preference model for recommendation[C]// Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery ＆ Data Mining. New York:ACM Press, 2019: 1891-1899.
[24]	秦川, 祝恒书, 庄福振 ,等. 基于知识图谱的推荐系统研究综述[J]. 中国科学:信息科学, 2020,50(7): 937-956.
	QIN C , ZHU H S , ZHUANG F Z ,et al. A survey on knowledge graph-based recommender systems[J]. SCIENTIA SINICA Informationis, 2020,50(7): 937-956.
[25]	WANG H W , ZHANG F Z , WANG J L ,et al. RippleNet:propagating user preferences on the knowledge graph for recommender systems[C]// Proceedings of the 27th ACM International Conference on Information and Knowledge Management. New York:ACM Press, 2018: 417-426.
[26]	MA M , CAO Q , LIU X Y . A graph convolution collaborative filtering integrating social relations recommendation method[J]. Applied Sciences, 2022,12(22): 11653.
[27]	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. New York:ACM Press, 2017: 1341-1350.
[28]	WAKITA Y , OKU K , KAWAGOE K . Toward fashion-brand recommendation systems using deep-learning:preliminary analysis[J]. International Journal of Knowledge Engineering, 2016,2(3): 128-131.
[29]	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 on Conference on Information and Knowledge Management. New York:ACM Press, 2015: 1661-1670.

ID	年龄段
0	＜18
1	18～24
2	25～34
3	35～44
4	45～49
5	50～55
6	＞56

ID	名称
0	other or not specified（其他或未指定）
1	academic/educator（学术/教育工作者）
2	artist（艺术家）
3	clerical/admin（文书/行政）
4	college/grad student（大学/研究生）
5	customer service（客户服务）
6	doctor/health care（医生/医疗保健）
7	executive/managerial（执行/管理）
8	farmer（农民）
9	homemaker（操持家务者）
10	K-12 student（K-12学生）
11	lawyer（律师）
12	programmer（程序员）
13	retired（已退休人员）
14	sales/marketing（销售/营销）
15	scientist（科学家）
16	self-employed（个体经营者）
17	technician/engineer（技术员/工程师）
18	tradesman/craftsman（商人/工匠）
19	unemployed（失业人员）
20	writer（作家）

潜在特征向量维度/维	MLP0	MLP1	MLP2	MLP3	MLP4	MLP5	MLP6	MLP7	MLP8
10	0.708 4	0.578 9	0.785 4	0.569 6	0.804 3	0.742 0	0.857 9	0.949 8	0.871 2
20	0.734 4	0.685 1	0.828 3	0.844 7	0.892 4	0.948 0	0.956 1	0.975 8	0.979 1
30	0.691 0	0.769 0	0.678 9	0.921 8	0.927 3	0.953 8	0.968 7	0.980 4	0.977 6
40	0.710 2	0.594 3	0.800 0	0.886 6	0.943 4	0.971 5	0.964 4	0.980 3	0.975 8
50	0.688 8	0.776 0	0.889 9	0.922 4	0.974 8	0.977 6	0.980 5	0.983 1	0.980 0
60	0.706 0	0.596 2	0.879 0	0.974 2	0.970 2	0.976 1	0.980 2	0.978 9	0.980 6
70	0.735 6	0.659 6	0.958 9	0.973 6	0.972 4	0.981 4	0.982 1	0.978 9	0.980 3
80	0.726 7	0.615 4	0.890 9	0.952 8	0.974 5	0.979 9	0.978 4	0.981 1	0.983 4

潜在特征向量维度/维	MLP0	MLP1	MLP2	MLP3	MLP4	MLP5	MLP6	MLP7	MLP8
10	0.605 7	0.450 5	0.727 5	0.470 0	0.757 0	0.717 0	0.843 0	0.940 9	0.860 2
20	0.651 6	0.595 7	0.785 1	0.841 9	0.881 5	0.939 9	0.948 8	0.968 8	0.972 5
30	0.596 5	0.709 0	0.625 0	0.912 0	0.919 0	0.945 3	0.960 6	0.972 5	0.969 0
40	0.627 0	0.499 3	0.774 2	0.871 0	0.933 7	0.962 7	0.956 0	0.972 7	0.969 1
50	0.612 8	0.741 6	0.874 8	0.913 3	0.967 33	0.969 9	0.973 1	0.974 9	0.973 5
60	0.640 7	0.508 1	0.864 4	0.966 7	0.962 5	0.968 2	0.973 8	0.972 2	0.974 3
70	0.673 6	0.599 8	0.948 9	0.968 3	0.965 0	0.973 1	0.975 1	0.971 5	0.973 5
80	0.672 7	0.542 1	0.872 4	0.944 3	0.967 8	0.973 6	0.971 4	0.974 3	0.975 5

基于用户多特征融合的个性化推荐模型

Personalized recommendation model based on users multi-features fusion

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 12

参考文献 29

相关文章 15

Metrics

推荐阅读 0

[1]	诸葛斌, 尹正虎, 斯文学, 颜蕾, 董黎刚, 蒋献. 基于学生知识追踪的多指标习题推荐算法[J]. 电信科学, 2022, 38(9): 129-143.
[2]	周杰, Esono Mikue Bernardo Esono, 王学英, 周惠婷, 罗宏. 基于SLM-PTS算法融合的NC-OFDM峰均比优化[J]. 电信科学, 2022, 38(7): 63-74.
[3]	李攀攀, 谢正霞, 乐光学, 刘鑫. 基于深度学习的无线通信接收方法研究进展与趋势[J]. 电信科学, 2022, 38(2): 1-17.
[4]	申情, 郭文宾, 楼俊钢, 余强国. 考虑多层次潜在特征的个性化推荐模型[J]. 电信科学, 2022, 38(2): 71-83.
[5]	陈志宏, 王明晓. 计算机视觉在智慧安防中的应用[J]. 电信科学, 2021, 37(8): 142-147.
[6]	唐博恒, 柴鑫刚. 基于云边协同的计算机视觉推理机制[J]. 电信科学, 2021, 37(5): 72-81.
[7]	孙姝君, 彭盛亮, 姚育东, 杨喜. 基于深度学习的调制识别综述[J]. 电信科学, 2021, 37(5): 82-90.
[8]	彭双, 王晓东, 彭宗举, 陈芬. 基于深度学习的快速QTMT划分[J]. 电信科学, 2021, 37(4): 73-81.
[9]	胡道允, 齐进, 陆钱春, 李锋, 房红强. 基于深度学习的流量工程算法研究与应用[J]. 电信科学, 2021, 37(2): 107-114.
[10]	张捷, 杨丽花, 王增浩, 呼博, 聂倩. 一种新型的基于深度学习的时变信道预测方法[J]. 电信科学, 2021, 37(1): 39-47.
[11]	李远宁,宁柏锋,董召杰. 电网机巡图像分析框架与深度学习方法[J]. 电信科学, 2020, 36(8): 167-174.
[12]	张婷婷,章坚武,郭春生,陈华华,周迪,王延松,徐爱华. 基于深度学习的图像目标检测算法综述[J]. 电信科学, 2020, 36(7): 92-106.
[13]	郭锐,冉凡春. 基于卷积神经网络的极化码译码算法[J]. 电信科学, 2020, 36(6): 119-124.
[14]	汪少敏,王铮,任华. 利用深度学习融合模型提升文本内容安全的研究[J]. 电信科学, 2020, 36(5): 25-30.
[15]	盛剑涛,陈茂飞,刘东鑫,汪来富,史国水,金华敏. 基于组合分类器的恶意域名检测技术[J]. 电信科学, 2020, 36(5): 47-55.