论文部分内容阅读
协同过滤是当前应用在信息推荐系统中最成功的技术之一。但随着用户数量和所需过滤信息的增加,计算复杂度迅速增长,大多数推荐系统都因集中式的体系结构而面临可扩展性差的问题。本文提出了一种基于非结构化P2P网络的协同过滤推荐机制,采用基于词汇链的方法构建资源对象描述向量,建立由偏好资源对象集合构成的用户模型,并且根据用户的兴趣变化,通过动态邻居重组的方法获得实时的个性化推荐。实验数据表明采用基于非结构化P2P网络的协同过滤推荐机制较传统集中式推荐方案有更好的可扩展性和预测准确性。
Collaborative filtering is one of the most successful techniques currently applied in information recommendation systems. However, with the increase in the number of users and the required filtering information, the computational complexity increases rapidly. Most recommended systems are faced with the problem of poor scalability due to a centralized architecture. This paper presents a recommendation mechanism of collaborative filtering based on unstructured P2P networks. The vocabulary description method is used to construct resource object descriptor vector, and a user model composed of preference resource objects is established. Based on the change of interest of users, Reorganized method to get real-time personalized recommendations. Experimental data show that the proposed collaborative filtering mechanism based on unstructured P2P networks has better scalability and predictive accuracy than traditional centralized recommender schemes.