网站首页  |  期刊介绍  |  编委会  |  投稿指南  |  在线订阅  |  联系我们English
何利,胡飘.基于用户多维度信任的冷启动推荐模型[J].重庆邮电大学学报(自然科学版),2018,30(6):827-834. 本文二维码信息
二维码(扫一下试试看!)
基于用户多维度信任的冷启动推荐模型
Cold start recommendation model based on user multi-dimension trust
投稿时间:2017-09-04  修订日期:2018-03-20
DOI: 10.3979/j.issn.1673-825X.2018.06.014
中文关键词:  协同过滤  社交信任  多维度信任  冷启动
English Keywords:collaborative filtering  social trust  multi dimension trust  cold start
基金项目:国家自然科学基金(61602073)
作者单位E-mail
何利 重庆邮电大学 计算机科学与技术学院, 重庆 400065 cquptheli@126.com 
胡飘 重庆邮电大学 计算机科学与技术学院, 重庆 400065 2451938514@qq.com 
摘要点击次数: 166
全文下载次数: 80
中文摘要:
      在传统协同过滤(collaborative filtering,CF)算法中存在着用户冷启动低效推荐问题,基于社交信息的社会化推荐算法通过引入用户的社交关系来缓解冷用户的数据稀疏问题,具有很好的研究应用前景。但该算法对用户间的信任度量较为片面单一而难以准确地对冷用户做出个性化推荐。针对此缺点,从不同维度量化分析了影响用户信任的因素,理论推导出用户多维度信任度量模型,将该模型计算得到的用户综合信任与传统协同过滤中的用户评分相似度值进行有效线性融合,提出了一种基于用户多维度信任的冷启动推荐模型。通过使用真实数据集Epinions并采用留一法进行实验对比分析。实验结果表明,提出的模型在对冷启动用户的推荐中,其平均绝对误差(mean absolute error,MAE)、覆盖率(rating coverage,RC)和F1值(F-measure)3个评估指标相比其他算法有了明显改善。
English Summary:
      In the traditional collaborative filtering algorithm, there is the problem of low inefficient recommendation on user cold start, and the recommendation algorithm based on social information can alleviate the data sparseness problem by introducing users’ social relationships. This algorithm has good prospects of application. However, because of the trust measure between users is one-sided, These algorithms are difficult to make personalized recommendation for cold users. In order to solve the problem, this paper firstly analyzes the factors of user trust from different dimensions. Especially, a multi-dimensions trust model is obtained to measure comprehensive user trust by integrating these factors. In addition, the user comprehensive trust and the rating similarity are effectively merged to replace the user similarity of the traditional collaborative filtering. Thus a cold start recommendation model based on user multi-dimension trust is proposed in this paper. Compared with other algorithms based on user or trust, the experimental results show that, for cold start users, the proposed model significantly improves the mean absolute error (MAE), rating coverage (RC) and F1 (F-Measure).
HTML    PDF浏览   查看/发表评论  下载PDF阅读器
版权所有 © 2009 重庆邮电大学期刊社  
地址:重庆市 南岸区 重庆邮电大学 期刊社 邮编:400065
电话:023-62461032 E-mail : journal@cqupt.edu.cn
meinv 海贼王论坛