A Self-Organized Personalized Recommendation System Based on Peer-to-Peer Networks.pdf

心智与计算 —————————————————————————————————————————————————————————————— 147 心智与计算, Volume 1 (2007), 147 -153 文章编号：MC - 2007-012 收稿日期：2006-09-30 出版日期：2007-01 ? 2007 MC– 厦门大学信息与技术学院 A Self-Organized Personalized Recommendation System Based on Peer-to-Peer Networks GUO Feng , LIN Ying * (Dept. of Computer Science, Xiamen University, Xiamen 361005, China) linying@ Abstract：Personalized recommendation systems can help people to find interesting things and they are widely used in the world. Hybrid peer-to-peer recommendation systems intend to resolve problems of C/S recommendation systems by distributing items and calculating missions to all users. This paper presents our recent research work on the new self-organized personalized recommendation system based on pure peer-to- peer network, it is also a multi-agent system, which contains lots of agents work in their peer, share and recommend automatic documents with the other agents. From various perspectives, our work focuses on how to adapt the system for a hybrid pure P2P network and implement users’ inside actions. Several models are proposed to make our system recommending effectively. These models are then evaluated and validated through implements and analyses. The results show some advantages of the proposed approach for the hybrid filtering based on recommendation threshold instead of Top-N and recommendation with Authority. Key words：Personalized Recommendation System； Hybrid Filtering； P2P； Multi-Agent 基于 P2P 网络的自组织个性化推荐系统 郭 锋 , 林 颖* (厦门大学计算机科学系, 福建 厦门 361005) linying@ 摘要：个性化推荐系统可以帮助人们找到感兴趣的内容，这些系统正在全世界内被广泛使用着。 混和式的 P2P 推荐系统通过将共享的项目和计算任务分布到各个用户，从而能够解决基于 C/S 的推荐 系统中存在的一些问题。本文描述了我们最近研究的基于 P2P 网络的自组织文章推荐系统，它同时也 是一个多 Agent 系统，该系统内将每个 Agent 作为一个对等点，能够保存文档，并自动向其他的用户共 享和推荐这些文档。我们主要研究了如何将推荐系统改造成适合混和式的 P2P 网络，以及如何能够迎 合用户的内子需求。我们提出了多个模型来提高我们系统的推荐性能，并通过仿真实验来评价和分析 这些模型。实验结果显示使用推荐阈值代替 Top-N 阈值以及引入推荐权威度能够带来系统性能上的很 大提高。 关键词： 个性化推荐系统；混和过滤；P2P；多 Agent 中图分类号：TP 393.0 文献标识码：A