摘要
协同过滤算法中最重要的一步是寻找相似用户,但用户评分数据的稀疏以及不诚实用户等问题,使仅仅依赖用户评分数据的传统协同过滤方法寻找的相似用户不够准确。在改进的基于用户数据的推荐算法中,用到用户评分数据和用户信息两种用户数据,通过对用户信息进行量化,得到用户信息矩阵。分别使用用户商品评分矩阵和用户信息矩阵来计算用户相似值,通过综合两种相似值来计算得到相似用户,并且通过加权来修正不诚实用户问题,通过筛选推荐用户来解决用户数据稀疏性问题。实验结果表明该方法能够有效地提高推荐精度。
In collaborative filtering algorithm,looking for similar users is the most important step. But due to the problems such as the sparseness of user rating data and the dishonest users,traditional collaborative filtering methods which rely only on users rating data fail in looking for the users with enough similarity. In the improved user data-based recommendation algorithm,two kinds of data are acquired,the user rating data and the user information data. Through quantifying the user information we get user information matrix,then we calculate the user similarity values by using user commodity rating matrix and user information matrix respectively,we also get similar users through integrating these two kinds of similarity values,correct the dishonest user problem by weighting,and solve the user rating data sparseness problem through screening the recommendation users. Experimental results show that this method can effectively improve the accuracy of recommendation.
出处
《计算机应用与软件》
CSCD
北大核心
2014年第8期245-248,共4页
Computer Applications and Software
基金
国家自然科学基金项目(61262025
61262024)
云南省应用基础研究计划面上项目(2012FB118)
云南省教育厅科学研究基金项目(2012Y257)
云南省软件工程重点实验室开放基金项目(2011SE09)
关键词
协同过滤
用户信息
不诚实用户
稀疏性
Collaborative filtering User information Dishonest user Sparseness
