融合信任度值与半监督密度峰值聚类的改进协同过滤推荐算法  

Collaborative Filtering Recommendation Algorithm Combining Trust Value and Semi-supervised Density Peak Clustering

作　　者：李昆仑[1] 王萌萌 于志波 翟利娜 LI Kun-lun;WANG Meng-meng;YU Zhi-bo;ZHAI Li-na(Electronic Information Engineering College,Hebei Unversity,Baoding 071000,China)

机构地区：[1]河北大学电子信息工程学院,河北保定071000

出　　处：《小型微型计算机系统》2020年第8期1613-1619,共7页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61672205)资助。

摘　　要：协同过滤推荐算法是个性化推荐系统中研究最多且应用最广的推荐算法之一,针对传统的协同过滤推荐算法中存在的数据稀疏性问题,导致算法的推荐精度不准确和推荐效率低等现象,本文提出了一种融合信任度值与半监督密度峰值聚类的改进协同过滤推荐算法.该算法有以下三个方面改进:1)通过半监督密度峰值聚类,将相似性用户进行聚类,降低目标用户的相似度计算时间;2)加入信任权值,精确地计算用户之间的直接信任度值;3)引入等效电阻式的间接信任度值的计算方式,充分挖掘用户之间的隐式信任信息.本文在进行相关的理论分析,并在不同数据集上进行了相应的实验验证,均表明了本文算法的有效性.Collaborative filtering recommendation algorithm is one of the most researched and widely used recommendation algorithms in personalized recommendation systems.Aiming at the problem of data sparsity in traditional collaborative filtering recommendation algorithms,it leads to inaccurate recommendation accuracy and low recommendation efficiency.This paper proposes an improved collaborative filtering recommendation algorithm that combines trust values and semi-supervised density peak clustering.The algorithm has the following three improvements:1)clustering similar users through semi-supervised density peak clustering to reduce the similarity calculation time of the target user;2)adding trust weights to accurately calculate the user weight The direct trust value between the users;3)Introducing the calculation method of the equivalent resistance indirect trust value to fully mine the implicit trust information between users.In this paper,the relevant theoretical analysis is carried out,and corresponding experimental verification is performed on different data sets,which all show the effectiveness of the algorithm.

关 键 词：协同过滤 半监督密度峰值聚类 直接信任 间接信任 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]