摘要
协同过滤技术基于用户的评分历史预测用户对某一项目的评分。基于用户的协同过滤技术可以利用传统的欧氏距离发现与用户的兴趣相近的近邻。针对欧氏距离并不能很好地反应用户之间的近邻关系的问题,一种新颖的基于欧氏距离的最小最大距离的方法被提出,用来发现用户近邻,称之为流形近邻。实验结果表明,基于流形近邻的协同过滤框架(Collaborative Filtering based on Manifold Neighbors,MNCF)与目前的协同过滤算法相比,性能有一定的提高。
The collaborative filtering technology predicts an active user's rating to an item based on his historical interests. Traditional userbased collaborative filtering takes the Euclidean distances as input to find the neighbors of an active user. However,the Euclidean distance cannot reflect the relationships well. A novel method which takes min-max distance based on the Euclidean distances to find an active user's neighbors is introduced and called manifold neighbors. According to the results of experiments,MNCF can outperform state-of-the-art collaborative filtering algorithms to some extent.
出处
《微型机与应用》
2016年第3期78-80,91,共4页
Microcomputer & Its Applications
基金
国家自然科学基金(61572444)
关键词
流形近邻
距离空间
协同过滤
视觉距离
最小最大距离
推荐系统
manifold neighbor
metric space
collaborative filtering
vision distance
min-max distance
recommendation system