摘要： 在协同过滤推荐系统中，通过对稀疏评分矩阵进行填充，可以提高对用户相似度的度量效果和系统的推荐精度。不同填充方法对相似度计算结果的影响存在较大差异。为解决该问题，针对3类填充方法构建的评分数据集，以最近邻算法进行推荐，分析传统相似度和基于云模型的相似度经2种方法优化后的度量效果，分别为各填充方法选取最有效的相似度优化方案。

关键词: 协同过滤, 最近邻, 相似度, 云模型

Abstract: In collaborative filter recommendation systems, the performance of user similarity measuring can be improved by filling the sparse marking matrix. Different filling method has different effect on similarity calculation result. To resolve this problem, this paper makes recommendation by using nearest neighbor algorithm on marking sets constructed by three kinds of filling methods separately, analyzes the measure performance optimized by two methods of traditional similarity measures and the similarity based on cloud model, and selects the most effective similarity measure optimization scheme for each filling method.

Key words: collaborative filter, nearest neighbor, similarity degree, cloud model

中图分类号: 

• TP391