论文部分内容阅读
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。
Personalized recommendation systems can help people find content of interest and are being used throughout the world. The hybrid P2P recommendation system can solve some problems in the C / S-based recommendation system by distributing the shared projects and computing tasks to various users. This paper describes our recently researched self-organizing article recommendation system based on P2P network, which is also a multi-agent system. Each Agent is regarded as a peer in this system, which can save documents and automatically share with other users Recommend these documents. We mainly studied how to transform a recommender system into a hybrid P2P network and how to meet the needs of users. We present several models to improve the recommended performance of our system and evaluate and analyze these models through simulation experiments. The experimental results show that the use of recommended threshold instead of Top-N threshold and the introduction of recommended authority can bring about a great improvement in system performance.