一种面向用户偏好定向挖掘的协同过滤个性化推荐算法  

A Collaborative Filtering Personalized Recommendation Algorithm Through Directionally Mining Users' Preferences

作　　者：王伟军[1] 宋梅青 

机构地区：[1]华中师范大学信息管理学院武汉430079 [2]华中师范大学青少年网络心理与行为教育部重点实验室武汉430079

出　　处：《现代图书情报技术》2014年第6期25-32,共8页New Technology of Library and Information Service

基　　金：国家自然科学基金项目“基于用户偏好感知的SaaS服务选择优化研究”(项目编号:71271099);湖北省自然科学基金创新群体重点项目“基于云计算的知识集成与服务研究”(项目编号:2011CDA116)的研究成果之一

摘　　要：【目的】解决协同过滤推荐的可扩展性问题和数据稀疏性问题。【方法】提出一种面向用户偏好定向挖掘的协同过滤算法。该算法以时间为约束,第一阶段先寻找基于项目的弱相似用户;第二阶段基于用户关联性和属性相似性进行定向挖掘,形成推荐集合。【结果】实验结果表明,新算法的时间复杂度降低一个数量级,并且数据越稀疏,推荐精度的领先优势越大。【局限】该算法基于用户已表现出的偏好进行深度推荐,对未表现出的其他偏好暂未涉及。【结论】该算法在提升可扩展性的同时,对数据稀疏性也有很强的适应能力。[Objective] To solve the scalability problem and data sparsity problem of the collaborative filtering. [Methods] This paper proposes an algorithm of collaborative filtering personalized recommendation through directionally mining users＇ preferences. Introducing time as a variable, the algorithm excavates in two stages. The first stage is to find the project-based weak similar users, the second stage is to use users＇ relevance and attribute similarity so as to do directional excavation and form a collection of recommendation. [Results] Experimental results show that the time complexity of the new algorithm reduces a magnitude. Furthermore, the more sparser the data is, the greater leading advantage the recommendation accuracy has. [Limitations] The algorithm recommends deeply by analyzing the users＇ existed preferences, and it doesn＇t involve the users＇ preferences which haven＇t appeared. [Conclusions] This algorithm has a strong ability to adapt to data sparsity and enhances its scalability at the same time.

关 键 词：协同过滤 用户偏好 个性化推荐 推荐算法 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]