摘要：

协同过滤算法是个性化推荐系统中应用较广的算法之一。随着用户数量及项目数量的增加，数据的稀疏问题成为影响个性化推荐质量的重要因素。为此，提出一种基于相似度传递的协同过滤算法。该算法能使大于阈值的用户相似度在有限路径长度上传递，增加可用于计算推荐值的用户最近邻居的数量，减少数据稀疏问题的影响，提高推荐质量。

关键词: 协同过滤, 相似度传递, 数据稀疏, 平均绝对误差

Abstract:

Collaborative filtering is one of the most widely used algorithms in the personalized recommendation system. With the increase in the number of users and items, the data sparsity becomes the important factor which affects the quality of the personalized recommendation. To address this issue, a new collaborative filtering algorithm based on the similarity propagation is introduced. The similarity between users can propagate in a limited path length as long as the similarity greater than the threshold. So the number of the nearest neighbors which is used to prediction increases and the quality of the recommendation improved because of the relief of the data sparsity.

Key words: collaborative filtering, similarity propagation, data sparsity, Mean Absolute Error(MAE)

中图分类号: 

• TP311.52