电信科学 ›› 2015, Vol. 31 ›› Issue (6): 68-74.doi: 10.11959/j.issn.1000-0801.2015113

• 研究与开发 • 上一篇    下一篇

基于多社交数据源的协同推荐方法研究

王瑞琴1,潘俊2,李一啸3   

  1. 1 湖州师范学院信息与控制技术研究所 湖州 313001
    2 温州大学建模与数据挖掘研究所 温州 325035
    3 浙江财经大学信息学院 杭州 310018
  • 出版日期:2015-07-23 发布日期:2015-08-03
  • 基金资助:
    国家自然科学基金资助项目;国家自然科学基金资助项目;国家教育部科学基金资助项目;浙江省自然科学基金资助项目;浙江省科技计划基金资助项目;浙江省科技计划基金资助项目

Research on Collaborative Recommendation Method Based on Multiple Data Sources of Social Network

Ruiqin Wang1,Jun Pan2,Yixiao Li3   

  1. 1 Institute of Information and Control Technology,Huzhou University,Huzhou 313001,China
    2 Institute of Business Modeling and Data Mining,Wenzhou University,Wenzhou 325035,China
    3 School of Information,Zhejiang University of Finance and Economics,Hangzhou 310018,China
  • Online:2015-07-23 Published:2015-08-03
  • Supported by:
    The National Science Foundation of China;The National Science Foundation of China;Science Foundation of Ministry of Education of China;Zhejiang Provincial Natural Science Foundation;Zhejiang Provincial Technology Program;Zhejiang Provincial Technology Program

摘要:

协同过滤推荐作为一种有效的推荐方法,普遍存在数据稀疏性和冷启动问题,利用社交网络的多项数据源对协同推荐方法进行了改进。为了克服评分矩阵的稀疏性问题,提出结合用户评分相似度和用户信任度选择推荐邻居,同时对用户相似度计算进行了改进;提出了一种简单有效的信任推理方法,能够识别出用户间隐含的间接信任关系,进一步缓解了数据稀疏性问题;为了解决推荐系统的冷启动问题,提出综合利用项目的类型属性信息和领域专家信息进行联合推荐。实验结果表明,提出的改进策略非常有效,在精度和召回率方面都较已有方法具有明显改善。

关键词: 社交网络, 个性化推荐, 信任推理, 多数据源, 领域专家

Abstract:

As an effective recommendation method,collaborative filtering typically has the data sparsity and cold-start problems. It was proposed that using multiple data sources of social network to overcome the above problems. First of a11,both the rating similarity and the social trust between users were considered to resolve the data sparsity problem. Then a simple and effective trust reasoning method was proposed to identify the implicit trust relationship between users. In order to solve the cold-start problem,information of the category of items and domain experts was used for joint recommendation. Experimental results show that the proposed algorithm has significantly better precision and reca11 than existing methods.

Key words: social network, personalized recommendation, trust inference, multiple data source, domain expert

No Suggested Reading articles found!