基于用户特征的分步协同推荐算法

协同过滤是解决信息过载问题的一种有效技术。针对基于内存的推荐面临着可扩展性问题、基于模型的推荐需要训练大量参数的问题进行了研究,从而提出了基于用户特征的K-means用户聚类算法,然后用分步协同过滤框架融合基于项目和基于用户的协同过滤给每一个聚簇训练一个模型。实验结果表明,提出的算法能极大地提高推荐精度,同时在一定程度上解决了基于模型和基于内存推荐存在的不足。...

Full description

Saved in:
Bibliographic Details
Published in计算机应用研究 Vol. 34; no. 4; pp. 1047 - 1049
Main Author 黄文明 程广兵 邓珍荣 周先亭
Format Journal Article
LanguageChinese
Published 桂林电子科技大学广西可信软件重点实验室,广西桂林,541004%桂林电子科技大学计算机科学与工程学院,广西桂林,541004 2017
Subjects
分步协同过滤框架
协同过滤
模型
用户特征
聚类算法
collaborative filtering
distributed collaborative filtering framework
clustering algorithm
聚类算法
协同过滤
用户特征
模型
user characteristics
model
分步协同过滤框架
Online AccessGet full text
ISSN1001-3695
DOI10.3969/j.issn.1001-3695.2017.04.020

Cover

More Information
Summary:协同过滤是解决信息过载问题的一种有效技术。针对基于内存的推荐面临着可扩展性问题、基于模型的推荐需要训练大量参数的问题进行了研究,从而提出了基于用户特征的K-means用户聚类算法,然后用分步协同过滤框架融合基于项目和基于用户的协同过滤给每一个聚簇训练一个模型。实验结果表明,提出的算法能极大地提高推荐精度,同时在一定程度上解决了基于模型和基于内存推荐存在的不足。
Bibliography:51-1196/TP
Collaborative filtering was an effective technique addressing the information overload problem. Aiming at the memory based recommendation suffered from difficulty in scalability, model-based recommendation had a multitude of parameters to train conducted a study, thus this paper proposed K-means user clustering algorithm based on user characteristics. Then, by using distributed collaborative filtering framework mixed with item based and user based collaborative filtering to train a model to every cluster. The experimental result shows that the proposed algorithm can greatly improve the recommendation precision. And solves the limitations of model based and memory based recommendations to a certain extent at the same time.
collaborative filtering; user characteristics; clustering algorithm; distributed collaborative filtering framework; model
Huang Wenminga, Cheng Guangbingb, Deng Zhenronga, Zhou Xiantingb （ a. Guangxi Key Laboratory of Trusted Software, b. School of Computer Science ＆ Engineering, Gui
ISSN:1001-3695
DOI:10.3969/j.issn.1001-3695.2017.04.020