华东师范大学学报(自然科学版) ›› 2019, Vol. 2019 ›› Issue (3): 63-77.doi: 10.3969/j.issn.1000-5641.2019.03.008

• 计算机科学 • 上一篇    下一篇

基于反向最远邻的商品推荐算法研究

王凯1, 李博涵1,2,3, 万朔1, 张安曼1, 关东海1   

  1. 1. 南京航空航天大学 计算机科学与技术学院, 南京 211106;
    2. 软件新技术与产业化协同创新中心, 南京 211106;
    3. 江苏易图地理信息科技股份有限公司, 江苏 扬州 225000
  • 收稿日期:2018-08-03 出版日期:2019-05-25 发布日期:2019-05-30
  • 通讯作者: 李博涵,男,博士,副教授,硕士生导师,研究方向为空间与时空数据库、地理信息系统等.E-mail:bhli@nuaa.edu.cn. E-mail:bhli@nuaa.edu.cn
  • 作者简介:王凯,男,硕士研究生,主要研究方向为时空数据库、推荐系统.E-mail:hxyrqx123@sina.com.
  • 基金资助:
    国家自然科学基金(61728204,61672284,41301407);南京航空航天大学科研基地创新基金(NJ20160028);南京航空航天大学青年科技创新基金(NT2018028);智能电网保护和运行控制国家重点实验室基金(国防预研领域);江苏省高校优势学科建设工程资助项目;高安全系统的软件开发与验证技术工业和信息化部重点实验室项目(NJ2018014)

Research on a commodity recommendation algorithm based on reverse furthest neighbor

WANG Kai1, LI Bo-han1,2,3, WAN Shuo1, ZHANG An-man1, GUAN Dong-hai1   

  1. 1. College of Computer Science and Technology, Nanjing University of Aeronautics and Astronautics, Nanjing 211106, China;
    2. Collaborative Innovation Center of Novel Software Technology and Industrialization, Nanjing 211106, China;
    3. Jiangsu Easymap Geographic Information Technology Corp. Ltd, Yangzhou Jiangsu 225000, China
  • Received:2018-08-03 Online:2019-05-25 Published:2019-05-30

摘要: 有效的推荐算法可以最大限度地发掘商品的价值.通过研究用户的偏好,分析了从海量商品信息中为用户推荐感兴趣内容的方法.目前大多数推荐系统向用户推荐的是较为流行的商品,而忽略了那些当下不"热门",却有着巨大潜力的商品.以发掘小众中的大众商品为目的,提出了一种基于反向最远邻(ReverseFurthest Neighbor,RFN)查询的商品推荐算法:基于专家用户的信任协同过滤算法,替代传统用户相似匹配的协同过滤推荐算法;利用幂律对商品进行范围缩减,优化系统筛选的效率,实现了对有潜在价值商品的推荐,使小众商品属性的分布得到更深层次的挖掘.实验结果表明本文推荐算法输出结果质量较高,适用于解决部分"长尾问题".

关键词: 推荐系统, 协同过滤, 反向最远邻, 幂律

Abstract: Effective recommendation algorithms can help maximize the value of a product. By studying the user's preferences, we can recommend underlying contents for the user from mass merchandise information. However, most recommendation systems focus on popular products, ignoring those products that are currently not popular but have huge potential. Our recommendation system, based on reverse furthest neighbor (RFN) queries, uses the idea of mining popular products in niche markets. We improve the traditional collaborative filtering recommendation algorithm and adopt a collaborative filtering algorithm based on expert users. The modified algorithm can recommend products with potential value based on the power law, which makes the distribution of minority mined products more visible to users. Experimental results show that the quality of the proposed algorithm is high and is suitable for partially addressing the long tail problem.

Key words: recommendation system, collaborative filtering, reverse furthest neighbor (RFN), power law

中图分类号: