基于上下文的分布式协同过滤推荐技术

Context-based Distributed Collaborative Filtering Recommendation Technology

针对传统协同过滤推荐技术应用于大规模动态数据集时难以兼顾准确度和效率的问题,提出一种基于上下文的分布式协同过滤推荐技术,引入推荐上下文的概念,并在此基础上充分考虑用户的即时兴趣以提高推荐的准确度,采用评分矩阵的分布式存储和计算以提高推荐的效率。实验结果表明,该分布式协同过滤技术能同时保证推荐的准确度和效率,使其在大规模动态数据集上的应用更具优势。

Aiming at the problems that the traditional Collaborative Filtering（CF） technology can hardly take both accuracy and efficiency into consideration when applied to large-scaled dynamic datasets, a context-based distributed collaborative filtering recommendation technology is proposed. The concept of recommend context is introduced and then based on that concept, it can enhance the recommend accuracy by taking users＇ immediate interest into account, and improve the algorithm＇s efficiency through the distributed store and computation of the rating matrix. Experimental results show that the context-based distributed collaborative filtering recommendation technology can guarantee both recommend accuracy and efficiency, and gains more advantages when it is applied to large-scaled dynamic data sets.