基于深度学习的混合兴趣点推荐算法

冯浩; 黄坤; 李晶; 高榕; 刘东华; 宋成芳

doi:10.11999/JEIT180458

基于深度学习的混合兴趣点推荐算法

doi: 10.11999/JEIT180458

冯浩¹,
黄坤¹,
李晶^2, ,,
高榕²,
刘东华²,
宋成芳²

1.
中国舰船研究设计中心武汉 430064
2.
武汉大学计算机学院武汉 430072

基金项目: 国家自然科学基金(41201404)，中央高校基本科研业务费专项资金(2042015gf0009)

详细信息

作者简介:
冯浩：男，1979年生，博士，高级工程师，研究方向为体系结构和系统工程

黄坤：男，1979年生，博士，高级工程师，研究方向为人工智能和系统工程

李晶：男，1967年生，博士，教授，研究方向为数据挖掘和多媒体技术

高榕：男，1981年生，博士，研究方向为数据挖掘和智能推荐

刘东华：女，1989年生，博士生，研究方向为数据挖掘和智能推荐

宋成芳：男，1978年生，博士，讲师，研究方向为可视化分析和位置服务

通讯作者:
李晶　leejingcn@163.com

中图分类号: TP311
计量
- 文章访问数: 3622
- HTML全文浏览量: 1718
- PDF下载量: 203
- 被引次数: 8
出版历程
- 收稿日期: 2018-05-14
- 修回日期: 2018-11-26
- 网络出版日期: 2018-12-05
- 刊出日期: 2019-04-01

Hybrid Point of Interest Recommendation Algorithm Based on Deep Learning

1.
China Ship Development and Design Center, Wuhan 430064, China
2.
Computer School, Wuhan University, Wuhan 430072, China

Funds: The National Natural Science Foundation of China (41201404), The Fundamental Research Funds for the Central Universities of China (2042015gf0009)

摘要

摘要:
针对现有兴趣点推荐的初始化和忽视评论信息语义上下文信息的问题，将深度学习融入推荐系统中已经成为兴趣点推荐研究的热点之一。该文提出一种基于深度学习的混合兴趣点推荐模型(MFM-HNN)。该模型基于神经网络融合评论信息与用户签到信息来提高兴趣点推荐的性能。具体地，利用卷积神经网络学习评论信息的特征表示，利用降噪自动编码对用户签到信息进行初始化。进而，基于扩展的矩阵分解模型融合评论信息特征和用户签到信息的初始值进行兴趣点推荐。在真实签到数据集上进行实验，结果表明所提MFM-HNN模型相比其他先进的兴趣点推荐具有更好的推荐性能。
- 推荐算法 /
- 兴趣点 /
- 矩阵分解 /
- 神经网络 /
- 深度学习
Abstract:
When modeling user preferences, the current researches of recommendation ignore the problem of modeling initialization and the review information accompanied with rating information for recommender models, integrating deep learning into the recommendation system becomes a hotspot of Point-Of-Interest (POI) recommendation. In this paper, a new POI recommendation model called Matrix Factorization Model integrated with Hybrid Neural Networks (MFM-HNN) is proposed. The model improves the performance of POI recommendation by fusing review text and check-in information based on Neural Network (NN). Specifically, the convolutional neural network is used to learn the feature representation of the review text and the check-in information is initialized by using the stacked denoising autoencoder. Furthermore, the extended matrix factorization model is exploited to fuse the review information feature and the initial value of the check-in information for POI recommendation. As is shown in the experimental results on real datasets, the proposed MFM-HNN achieves better recommendation performances than the other state-of-the-art POI recommendation algorithms.
- Recommendation algorithm /
- Point-Of-Interest (POI) /
- Matrix factorization /
- Neural Network (NN) /
- Deep learning

HTML全文

图 1 基于混合神经网络矩阵分解的兴趣点推荐模型

下载: 全尺寸图片幻灯片

图 2 MFM-HNN模型基于LA数据集和NYC数据集与其他4个模型的推荐性能对比

下载: 全尺寸图片幻灯片

图 3 基于LA和NYC数据集的5个初始化方法的性能对比

下载: 全尺寸图片幻灯片

图 4 MFM-HNN模型基于LA数据集和NYC数据集在不同层数的性能对比

下载: 全尺寸图片幻灯片

表 1 MFM-HNN模型学习算法

　输入：

${{{x}}_i},{{{S}}^u},{{{S}}^v},{{\tilde{{S}}}^u},{{\tilde{{S}}}^v},T,B$

　输出：

${\cal{L}}$

　(1) For

$t <$ T Do

　(2) 从兴趣点评论中随机选取一个兴趣点的评论矩阵

${{{x}}_i}$ 进行训练，　　训练批次大小为

${\beta _0}$ ，每一个批次的大小为

$B$ ，计算训练过程　　中的损失

${{\cal{L}}_{\rm{cnn}}}$

　(3) if

$t >$ T or

${{\cal{L}}_{\rm{cnn}}}$ 足够小

　(4) end

　(5) for

$t <$ T Do

　(6) 从兴趣点评分中随机选取一个兴趣点的用户-兴趣点对

$({\tilde{{s}}}_i^u,{\tilde{{s}}}_i^v)$ 　　进行训练，训练批次大小为

${\beta _1}$ ，每一个批次的大小为

$B$ ，计算　　训练过程中的损失

${{\cal{L}}_{\rm{ui}}}$

　(7) if

$t >$ T or

${{\cal{L}}_{\rm{ui}}}$ 足够小

　(8) end

　(9) 计算最终的损失值

${\cal{L}} = {{\cal{L}}_{\rm{cnn}}} + {{\cal{L}}_{\rm{ui}}}$

　(10) return

${\cal{L}}$

下载: 导出CSV

表 2 数据集统计

数据统计	LA	NYC
用户数量	30,208	47,240
兴趣点数量	142,798	203,765
签到数量(评论)	244,861	388,954
用户-位置矩阵密度	5.68×10^–5	4.04×10^–5

下载: 导出CSV

参考文献(23)

LI Jun, BENEDIKTSSON J A, ZHANG Bing, et al. Spatial technology and social media in remote sensing: A survey[J]. Proceedings of the IEEE, 2017, 105(10): 1855–1864 doi: 10.1109/JPROC.2017.2729890

YANG, C, BAI Lanxiao, ZHANG Chao, et al. Bridging collaborative filtering and semi-supervised learning: A neural approach for POI recommendation[C]. Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Halifax, Canada, 2017: 1245–1254.

LIU Yiding, PHAM T A N, CONG Gao, et al. An experimental evaluation of point-of-interest recommendation in location-based social networks[J]. Proceedings of the VLDB Endowment, 2017, 10(10): 1010–1021 doi: 10.14778/3115404.3115407

MAZUMDAR P, PATRA B K, BABU K S, et al. Hidden location prediction using check-in patterns in location-based social networks[J]. Knowledge and Information Systems, 2017, 57(3): 571–601 doi: 10.1007/s10115-018-1170-5

余永红, 高阳, 王皓. 基于Ranking的泊松矩阵分解兴趣点推荐算法[J]. 计算机研究与发展, 2016, 53(8): 1651–1663 doi: 10.7544/issn1000-1239.2016.20160202

YU Yonghong, GAO Yang, and WANG Hao. A ranking based Poisson matrix factorization model for point-of-interest recommendation[J]. Journal of Computer Research and Development, 2016, 53(8): 1651–1663 doi: 10.7544/issn1000-1239.2016.20160202

LIM K H, CHAN J, LECKIE C, et al. Personalized trip recommendation for tourists based on user interests, points of interest visit durations and visit recency[J]. Knowledge and Information Systems, 2018, 54(2): 375–406 doi: 10.1007/s10115-017-1056-y

YIN Hongzhi, SUN Yizhou, CUI Bin, et al. LCARS: A location-content-aware recommender system[C]. Proceedings of the 19th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Chicago, USA, 2013: 221–229.

ZHANG Jiadong, CHOW Chiyin, and ZHENG Yu. ORec: An opinion-based point-of-interest recommendation framework[C]. Proceedings of the 24th ACM International on Conference on Information and Knowledge Management, Melbourne, Australia, 2015: 1641–1650.

WANG Hao, WANG Naiyan, and YEUNG D Y. Collaborative deep learning for recommender systems[C]. Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Sydney, Australia, 2015: 1235–1244.

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, Boston, USA, 2016: 233–240.

DONG Xing, YU Lei, WU Zhonghuo, et al. A hybrid collaborative filtering model with deep structure for recommender systems[C]. Proceedings of the 31st AAAI Conference on Artificial Intelligence, San Francisco, USA, 2017: 1309–1315.

任星怡, 宋美娜, 宋俊德. 基于位置社交网络的上下文感知的兴趣点推荐[J]. 计算机学报, 2017, 40(4): 824–841 doi: 10.11897/SP.J.1016.2017.00824

REN Xingyi, SONG Meina, and SONG Junde. Context-aware point-of-interest recommendation in location-based social networks[J]. Chinese Journal of Computer, 2017, 40(4): 824–841 doi: 10.11897/SP.J.1016.2017.00824

GAO Rong, LI Jing, LI Xuefei, et al. A personalized point-of-interest recommendation model via fusion of geo-social information[J]. Neurocomputing, 2018, 273: 159–170 doi: 10.1016/j.neucom.2017.08.020

LIAN Defu, ZHENG Kai, GE Yong, et al. GeoMF++: Scalable location recommendation via joint geographical modeling and matrix factorization[J]. ACM Transactions on Information Systems, 2018, 36(3): 33 doi: 10.1145/3182166

ZDUNEK R. Initialization of nonnegative matrix factorization with vertices of convex polytope[C]. Proceedings of the 11st International Conference on Artificial Intelligence and Soft Computing, Zakopane, Poland, 2012: 448–455.

YANG Cheng, SUN Maosong, ZHAO W X, et al. A neural network approach to jointly modeling social networks and mobile trajectories[J]. ACM Transactions on Information Systems, 2017, 35(4): 36 doi: 10.1145/3041658

DENG Shuiguang, HUANG Longtao, XU Guangdong, et al. On deep learning for trust-aware recommendations in social networks[J]. IEEE Transactions on Neural Networks and Learning Systems, 2017, 28(5): 1164–1177 doi: 10.1109/TNNLS.2016.2514368

GAO Huiji, TANG Jiliang, HU Xia, et al. Content-aware point of interest recommendation on location-based social networks[C]. Proceedings of the 29th AAAI Conference on Artificial Intelligence, Austin, Texas, 2015: 1721–1727.

ZHANG Fuzheng, YUAN N J, ZHENG Kai, et al. Exploiting dining preference for restaurant recommendation[C]. Proceedings of the 25th International Conference on World Wide Web, Montréal, Canada, 2016: 725–735.

LIU Yong, WEI Wei, SUN Aixin, et al. Exploiting geographical neighborhood characteristics for location recommendation[C]. Proceedings of the 23rd ACM International Conference on Information and Knowledge Management, Shanghai, China, 2014: 739–748.

LI Huayu, GE Yong, HONG Richang, et al. Point-of-interest recommendations: Learning potential check-ins from friends[C]. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, USA, 2016: 975–984.

ORCHAND M. Least square quantization in PCM[J]. IEEE Transaction on Information Theory, 1982, 28(2): 129–137 doi: 10.1109/TIT.1982.1056489

SHI J and MALIK J. Normalized cuts and image segmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2000, 22(8): 888–905 doi: 10.1109/34.868688

施引文献

期刊类型引用(6)

1.	李云飞，冯珂珂，张飞，陈斌，车延超，金岩. 基于频谱细化算法的超声多普勒流量计设计. 传感器与微系统. 2024(10): 79-83 . 百度学术
2.	邢涛，马春铭，冯亮，李爽，韦立登，李军. 一种用于距离维复杂空变多普勒SAR成像的改进运动补偿方法. 电子与信息学报. 2022(03): 1059-1066 . 本站查看
3.	姚山峰，贺青，欧阳鑫信，杨宇翔. 一种低轨双星高脉冲重复频率雷达信号的定位模糊消除算法. 电子与信息学报. 2021(03): 598-605 . 本站查看
4.	赵晨，乔钢，周锋. 基于正交移动双水下自主潜航器的水下合作目标定位方法. 电子与信息学报. 2021(03): 834-841 . 本站查看
5.	张轶，翟盛华，陶海红. 单星多波束天线下基于压缩感知的多目标干扰定位. 电子与信息学报. 2021(07): 1872-1878 . 本站查看
6.	曲志昱，王超然，孙萌. 基于改进迭代扩展卡尔曼滤波的3星时频差测向融合动目标跟踪方法. 电子与信息学报. 2021(10): 2871-2877 . 本站查看