摘要
传统协同过滤中用户相似性计算方法在数据稠密的情况下能得到比较可靠的结果.然而当数据稀疏时,相似度计算方法会受到数据稀疏性影响,使其计算结果不准确,特别是在冷启动用户中,相似度计算结果不可靠.为了解决以上问题,从用户社交关系的角度,提出一种新的基于社交关系的相似度传播式协同过滤推荐算法.首先对用户社交关系进行建模,量化用户之间社交关系;然后计算活跃用户之间的相似性得到可靠相似度;最后基于社交关系的可靠相似度传播来构建推荐方法.实验证明,提出的算法可以大幅度提高推荐精度,改善对冷启动用户的推荐.
In traditional collaborative filtering, user similarity calculation methods can get more reliable results when data is dense. While data is sparse, however,the similarity calculation methods are affected by the sparseness of data,which makes the results of these methods not accurate, especially in cold start users, the results of similarity calculation are not reliable. To solve above problems, a new collaborative filtering recommendation algorithm of similarity propagation based on social relation was proposed from the perspective of user social relation. Firstly, user social relation is modeled and social relationships between users are quantitated;then, the proposed algorithm calculates the similarity between active users and gets reliable similarity; finally, the recommendation method is built based on the reliable similarity propagation of social relation. Experiments show that the proposed algorithm can improve the accuracy of recommendation and the recommendation for cold users in a large scale.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第5期1073-1077,共5页
Journal of Chinese Computer Systems
基金
教育部规划基金项目(11YJA860028)资助
福建省自然科学基金项目(2013J01219)资助
关键词
社交关系
活跃用户
可靠相似度
相似度传播
协同过滤推荐
social relation
active user
reliable similarity
similarity propagation
collaborative filtering recommendation
