杨思渝,
左家莉,
王明文
江西师范大学计算机信息工程学院 南昌 330022
基金项目:国家自然科学基金(61462042, 61866018, 61876074)
详细信息
作者简介:叶继华:男,1966年生,教授,博士生导师,主要研究方向为智能信息处理、数据融合、物联网技术、图像处理
杨思渝:女,1994年生,硕士生,研究方向为智能信息处理
左家莉:女,1981年生,博士,副教授,硕士生导师,主要研究方向为自然语言处理
王明文:男,1964年生,教授,博士生导师,主要研究方向为自然语言处理、信息检索、机器学习、数据挖掘
通讯作者:叶继华 yjhwcL@163.com
中图分类号:TP391计量
文章访问数:153
HTML全文浏览量:110
PDF下载量:44
被引次数:0
出版历程
收稿日期:2020-05-07
修回日期:2021-04-26
网络出版日期:2021-06-02
刊出日期:2021-12-21
Research on POI Recommendation Model Based on Spatio-temporal Context Information
Jihua YE,,Siyu YANG,
Jiali ZUO,
Mingwen WANG
Institute of Computer Information Engineering, Jiangxi Normal University, Nanchang 330022, China
Funds:The National Natural Science Foundation of China (61462042, 61866018, 61876074)
摘要
摘要:随着基于位置的社交网络(LBSN)技术的快速发展,为移动用户提供个性化服务的兴趣点(POI)推荐成为关注重点。由于POI推荐面临着数据稀疏、影响因素多和用户偏好复杂的挑战,因此传统的POI推荐往往只考虑签到频率以及签到时间和地点对用户的影响,而忽略了签到序列中用户前后行为的关联影响。为了解决上述问题,该文通过序列的表示考虑签到数据的时间影响和空间影响,建立了时空上下文信息的POI推荐模型(STCPR),为POI推荐提供了更精准的个性化偏好。该模型基于序列到序列的框架下,将用户信息、POI信息、类别信息和时空上下文信息进行向量化后嵌入GRU网络中,同时利用了时间注意力机制、全局和局部的空间注意力机制来综合考虑用户偏好与变化趋势,从而向用户推荐感兴趣的Top-N的POI。该文通过在两个真实的数据集上实验来验证模型的性能。实验的结果表明,该文所提出的方法在召回率(Recall)和归一化折损累计增益(NDCG)方面优于几种现有的方法。
关键词:兴趣点推荐模型/
时空相关性/
上下文/
注意力机制
Abstract:With the high-speed development of Location-Based Social Networking (LBSN) technology, Point-Of-Interest(POI) recommendation for providing personalized services to mobile users has become the focus of attention. Because POI recommendation is faced with the challenges of sparse data, multiple influencing factors and complex user preferences, traditional POI recommendation usually only considers the influence of check-in frequency, check-in time and place on users, but ignores the correlation influence of users’ behaviors before and after the check-in sequence. In order to solve the above problems, this paper takes into account the time influence and spatial influence of the check-in data through the representation of the sequence, and establishes a Spatio-Temporal Context information of POI Recommendations (STCPR), provides a more accurate and personalized preference for POI Recommendations. The model based on the framework of sequence to sequence, the user information, POI information, categories and space-time context information are embedded in GRU helped after vectorization network, at the same time the time attention mechanism, the mechanism of spatial attention of the global and local comprehensive are used for consideration of user preferences and trends, and in Top - N POI is recommended to users. In order to verify the performance of the model, experiments on two real data sets show that the proposed method is superior to several existing methods in terms of recall rate (Recall) and Normalized Discounted Cumulative Gain (NDCG).
Key words:Point-Of-Interest (POI) recommendation model/
Spatiotemporal correlation/
Context/
Attention mechanism
PDF全文下载地址:
https://jeit.ac.cn/article/exportPdf?id=d12398e1-5929-449a-a841-c2755cd0db70