Abstract: Collaborative filtering is the most prevalent algorithm of personalized service，but there are always two difficult problems for collaborative filtering algorithms，that is，data sparsity and expansibility.In this paper，a collaborative filtering algorithm is proposed based on item category.On the foundation of classifying items，it converts rating matrix of user-item into user-category of item.And then it clusters users，finds the nearest neighbors of active users in the sub-clustering that the active users exist in.Finally，it recommends to active users.Experiments show that the proposed algorithm reduces the data sparsity and dimensionality， its recommendations are good，and the simultaneity and expansibility of the algorithm is improved effectively.

摘要： 协作过滤是当今应用最为普遍的个性化推荐算法，然而数据的稀疏性和算法的可扩展性一直是协作过滤算法所面临的两大问题。提出了一种新的推荐算法——基于资源的协作过滤算法。该算法在对资源项目依内容划分的基础上，将用户—项目评分矩阵转换为用户—资源类别评分矩阵，然后对用户聚类，在目标用户所在的簇中寻找其最近邻居并产生推荐。实验表明，该算法不仅降低了数据的稀疏性和维度，缩小了目标用户最近邻的查找范围，算法的扩展性得到了有效改善，而且提高了最近邻的准确度，推荐精度较以往传统算法有明显提高。