基于Mahout分布式协同过滤推荐算法分析与实现  

Analysis and Implementation of Distributed Collaborative Filtering Recommendation Algorithm Based on Mahout

作　　者：曾志浩[1] 张琼林 姚贝[1] 孙琪[1] 

机构地区：[1]湖南工业大学计算机与通信学院,湖南株洲412007

出　　处：《计算技术与自动化》2015年第3期67-72,共6页Computing Technology and Automation

基　　金：科技部科技支撑计划课题项目(2013BAJ10B14-5);湖南省自然科学基金面上课题项目(14JJ2115);湖南省教育厅科研重点项目(14A037)

摘　　要：随着信息技术和互联网的发展,在信息过载的时代,用户面对海量的信息,难以正确选择。协同过滤推荐是个性化推荐中比较成熟的算法,但其稀疏性、冷启动、可扩展性问题仍然存在,尤其是不能应用于分布式推荐。在Hadoop平台上,Mahout实现了分布式基于项目的协同过滤推荐算法,该算法能够有效解决传统算法的海量数据处理的效率问题和可扩展性问题。实验结果表明,Mahout上基于项目的协同过滤推荐算法具有较好的计算高效性和可扩展性。With the development of information technology and Internet,facing the vast amount of information,it is dif-ficult to correctly choose for users in the era of information overload.Collaborative filtering is a relatively mature algorithm in personalized recommendation,but its sparsity,cold start,scalability problems still exist,especially can not be applied to distributed recommendation.On the platform of Hadoop,Mahout realized the distributed item-based collaborative filtering recommendation algorithm,and the algorithm can effectively solve massive data processing efficiency and scalability problem of the traditional algorithm.The experimental results show that,collaborative filtering algorithm has the high calculation ef-ficiency and good scalability based on Mahout.

关 键 词：分布式协同过滤 Mahout 推荐系统 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]