ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2014, Vol. 51 ›› Issue (10): 2255-2269.doi: 10.7544/issn1000-1239.2014.20130056

• 信息处理 • 上一篇    下一篇

云计算环境下基于协同过滤的个性化推荐机制

朱夏,宋爱波,东方,罗军舟   

  1. (东南大学计算机科学与工程学院 南京 211189) (xzhu@seu.edu.cn)
  • 出版日期: 2014-10-01
  • 基金资助: 
    国家自然科学基金项目(61173065,61202297,61202299);北京市自然科学基金重点项目(4111003)

A Collaborative Filtering Recommendation Mechanism for Cloud Computing

Zhu Xia, Song Aibo, Dong Fang, Luo Junzhou   

  1. (School of Computer Science and Engineering, Southeast University, Nanjing 211189)
  • Online: 2014-10-01

摘要: 随着云计算时代的到来,应用数据量剧增,个性化推荐技术日趋重要.然而由于云计算的超大规模以及分布式处理架构等特点,将传统的推荐技术直接应用到云计算环境时会面临推荐精度低、推荐时延长以及网络开销大等问题,导致推荐性能急剧下降.针对上述问题,提出一种云计算环境下基于协同过滤的个性化推荐机制RAC.该机制首先制定分布式评分管理策略,通过定义候选邻居(candidate neighbor, CN)的概念筛选对推荐结果影响较大的项目集,并构建基于分布式存储系统的2个阶段评分索引,保证推荐机制快速准确地定位候选邻居;在此基础上提出基于候选邻居的协同过滤推荐算法(candidate neighbor-based distribited collaborative filtering algorithm, CN-DCFA),在候选邻居中搜索目标用户已评分项目的k近邻,预测目标用户的推荐集top-N.实验结果表明,在云计算环境下RAC拥有良好的推荐精度和推荐效率.

关键词: 协同过滤, 个性化推荐, 云计算, 候选邻居, 分布式评分管理

Abstract: The advent of cloud computing has witnessed a sharp augmentation in the amount of application data, the result of which gives more and more prominence to a personalized recommendation technique, which has been used by different kinds of Web applications. However, traditional collaborative filtering (CF) techniques, when applied to cloud computing which features a huge scale and distributed processing architecture, are confronted with some problems, such as low recommendation accuracy, long recommendation time and high network traffic. Consequently, the performance of recommendation mechanisms degrades drastically. To address this issue, a CF recommendation mechanism for cloud computing (RAC) is proposed in this paper. RAC first employs a candidate neighbor, the definition of which will also be given, to screen those items which exert great influence on the recommendation results. Then a 2-stage rating indexing structure based on distributed storage system will be constructed to ensure the quick and precise location of CN by RAC. On this basis, a CN-based CF recommendation algorithm CN-DCFA which searches k nearest neighbors between candidate neighbors can be provided. Meanwhile, CN-DCFA utilizes similarity to predict the recommendation results top-N of active users. Experiments have showed that our recommendation mechanism RAC, enjoying great efficiency and exactitude, is worthy of recommendation.

Key words: collaborative filtering, personalized recommendation, cloud computing, candidate neighbor, distributed rating management

中图分类号: