一种融合隐式信任的协同过滤推荐算法  

Research on Implied-trust Aware Collaborative Filtering Recommendation Algorithm

作　　者：陆坤[1] 谢玲[1] 李明楚[1] 

机构地区：[1]大连理工大学软件学院,辽宁大连116600

出　　处：《小型微型计算机系统》2016年第2期241-245,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61100194;61272173)资助

摘　　要：大数据背景下,如何从海量数据中挖掘出有用信息,从而提供精确及时的个性化推荐,引起了学者们的广泛兴趣.提出一种融合潜在的信任关系的基于协同过滤的个性化推荐算法,该算法从海量用户历史数据中,挖掘出潜在的信任关系,该信任关系涵盖了用户本身的活跃和用户间的喜好差异,可进行更精确的个性化推荐.仿真实验中,选定余弦相似度作为本文相似性的度量标准,同时对相似性和调和信任度进行调和平均,通过协同过滤算法得到个性化得推荐结果.在测试数据集上的大量实验证明,本文的推荐算法其准确率和多样性上优于传统算法,可保证推荐结果的准确全面.In big data era,it is important to provide users with more accurate recommendations. The paper proposes a trust-based col- laborative filtering recommendation algorithm which can explore potential trust relationship from huge of users＇ history data. The trust relation contains user activeness in the system and the similarity of users, so more accuracy personal recommendation result can be pro- duced. In simulation experiments, the paper uses cosine similarity as the standard of similarity evaluating, harmonic weight is computed according to the similarities and trust value and used to replace the weight of the traditional collaborative filtering algorithm. The exper- imental results show that the algorithm accuracy and diversity are superior to the traditional algorithm.

关 键 词：协同过滤推荐 隐式信任 推荐系统 信任机制 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]