电子学报 ›› 2017, Vol. 45 ›› Issue (6): 1530-1536.DOI: 10.3969/j.issn.0372-2112.2017.06.036

• 科研通信 • 上一篇    下一篇

基于上下文相似度和社会网络的移动服务推荐方法

俞春花1, 刘学军1,2, 李斌1, 章玮3   

  1. 1. 南京工业大学计算机科学与技术学院, 江苏南京 211816;
    2. 复旦大学, 上海 200433;
    3. 中国人民解放军 73677部队, 江苏南京 210016
  • 收稿日期:2015-09-15 修回日期:2016-02-03 出版日期:2017-06-25
    • 作者简介:
    • 俞春花 女,1991年1月出生于江苏省泰州市.现为南京工业大学计算机科学与技术学院硕士研究生.主要研究方向为数据挖掘和个性化推荐.E-mail:chunhuayu1991@163.com;刘学军 男,1970年6月出生于吉林省长春市.博士,现为南京工业大学教授、硕士生导师,CCF高级会员.主要研究方向为数据库、数据挖掘、传感器网络、隐私保护等.E-mail:lxj_njgd@163.com
    • 基金资助:
    • 国家自然科学基金 (No.61203072); 江苏省重点研发计划 (社会发展) (No.BE2015697)

Mobile Service Recommendation Based on Context Similarity and Social Network

YU Chun-hua1, LIU Xue-jun1,2, LI Bin1, ZHANG Wei3   

  1. 1. College of Computer Science and Technology, Nanjing Tech University, Nanjing, Jiangsu 211816, China;
    2. Fudan University, Shanghai 200433, China;
    3. 73677 PLA Troops, Nanjing, Jiangsu 210016, China
  • Received:2015-09-15 Revised:2016-02-03 Online:2017-06-25 Published:2017-06-25
    • Supported by:
    • National Natural Science Foundation of China (No.61203072); Key Research and Development Project of Jiangsu Province  (Social Development) (No.BE2015697)

摘要:

针对传统的基于协同过滤的移动服务推荐方法存在的数据稀疏性和用户冷启动问题,提出一种基于上下文相似度和社会网络的移动服务推荐方法(Context-similarity and Social-network based Mobile Service Recommendation,CSMSR).该方法将基于用户的上下文相似度引入个性化服务推荐过程,并挖掘由移动用户虚拟交互构成的社会关系网络,按照信任度选取信任用户;然后结合基于用户评分相似度计算发现的近邻,分别从相似用户和信任用户中选择相应的邻居用户,对目标用户进行偏好预测和推荐.实验表明,与已有的服务推荐方法TNCF、SRMTC及CF-DNC相比,CSMSR方法有效地缓解数据稀疏性并提高推荐准确率,有利于发现用户感兴趣的服务,提升用户个性化服务体验.

关键词: 移动服务推荐, 上下文, 相似度计算, 社会网络, 协同过滤, 稀疏性, 冷启动问题

Abstract:

Concentrating on the data sparsity problem and the new user cold-start problem faced by traditional collaborative filtering algorithm in mobile recommender system,an approach named CSMSR (Context-similarity and Social-network based Mobile Service Recommendation) is proposed.The approach integrates mobile users' context information and social network information into collaborative filtering recommendation process.Firstly,it imports the user-based context similarities into the personalized service recommendation process.Secondly,it searches the corresponding nearest neighbors for each mobile user according to the given mobile users' ratings and the mining social network.Finally,it predicts unknown mobile users' preferences and generates recommendations.The experimental results show that CSMSR outperforms the existing service recommendation methods,such as TNCF,SRMTC and CF-DNC in terms of MAE (Mean Absolute Error) and P@N,and it performs well in finding out the interested services of users and enhancing the user experience.

Key words: mobile service recommendation, context, similarity measure, social network, collaborative filtering, data sparsity, cold-start problem

中图分类号: