基于Spark技术的协同过滤推荐算法的可扩放性研究  

Research on scalability of collaborative filtering recommendation algorithm based on Spark technology

作　　者：祝永志 Zhu Yong-zhi(School of Information Science and Engineering,Qufu Normal University,Rizhao 276826,Chin)

机构地区：[1]曲阜师范大学信息科学与工程学院,日照276826

出　　处：《电子技术（上海）》2018年第7期41-42,40,共3页Electronic Technology

基　　金：山东省自然科学基金(ZR2013FL015);山东省研究生教育创新资助计划(SDYY12060)

摘　　要：传统的基于项目相似度的协同过滤算法在处理大数据集时会出现延迟过长及效率不高等问题。本文基于Hadoop分布式框架及Spark分布式并行编程模型,设计了一个高效的基于项目相似度的协同过滤算法。经实验验证,文章算法具有很好的可扩放性及很好的推荐精度。The traditional collaborative filtering algorithm based on item similarity will be time-consuming and inefficient when dealing with large data sets. Based on Hadoop distributed framework and Spark distributed parallel programming model, this paper designs an efficient collaborative filtering algorithm based on item similarity. Experimental results show that the algorithm has good scalability and good recommendation accuracy.

关 键 词：SPARK 协同过滤 HADOOP 可扩放性 推荐精度 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]