一种协同过滤方法及其在信息推荐系统中的实现

A Collaborative Information Filtering Method and Its Application in Information Recommending System

本文提出了一种基于模糊聚类技术的协同过滤方法,应用模糊聚类技术从项目的属性特征上对项目进行聚类,用隶属度的值来表示项目属于每个模糊簇的程度,由用户-项评分矩阵和模糊簇的隶属度值,构造用户-模糊簇的偏好矩阵,进而利用用户-模糊簇偏好矩阵获得用户相似群体,为用户实现基于协同过滤的文档推荐.最后,利用过滤技术实现了一个科技文献推荐系统,对多种推荐策略进行了验证.

In this paper the research of information filtering is firstly reviewed. A fuzzy-cluster-based collaboration filtering algorithm is proposed. In the algorithm the fuzzy-clustering method is applied to cluster the items characters. The grade of membership is used to describe which group the item belongs to. The User-Cluster profile matrix is composed of the User-ltem profile matrix and the grade of membership. The documents recommending are made based on the User-Cluster profile matrix. Finally one science document recommending system is designed using filtering methods. In the system multi-filtering methods are used and tested.