评分可信度条件下的协同过滤模型  

Collaborative Filtering Models with the Credibility of Ratings

作　　者：杨兴耀[1] 于炯[1] 吐尔根.依布拉音 英昌甜[1] 闫歌[1] 

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046

出　　处：《微电子学与计算机》2013年第10期25-28,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(61063042;61262088;61063026);新疆大学博士创新项目基金(XJUBSCX-2011007);新疆自治区自然科学基金项目(2011211A011)

摘　　要：通过对用户信任度进行量化,再从项目的角度进行调整,最终获得了较为准确的评分可信度度量.在此基础上,建立了评分可信度矩阵,并对提出的四种可信度相似性模型进行了优化.实验比较结果表明,基于不同的数据集,新提出的相似性模型在合理的时间开销下,相对于传统模型在项目预测准确性方面拥有出色的表现.This paper quantifies the degree of user credit and adjusts it by specific items, and then gains accurate quantification on the credibility of ratings. In the case, the paper establishes the credibility matrix of ratings, and proposes four optimized similarity models based on credibility. The comparative experimental results, tested on different data sets, demonstrate that the newly proposed models have better performance in prediction accuracy of items at a reasonable cost in time, compared to traditional models.

关 键 词：推荐系统 协同过滤 可信度 相似性模型 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]