摘要
随着大数据时代的到来,应用数据量剧增,个性化推荐技术日趋重要。传统的推荐技术直接应用于大数据环境时会面临推荐精度低、推荐时延长以及网络开销大等问题,导致推荐性能急剧下降。针对上述问题,提出用户共现矩阵乘子推荐策略,将用户相似度矩阵与项目评分矩阵相乘得到用户对项目的预测评分矩阵,从而生成对每个用户的候选推荐项目集;在此基础上,根据分布式处理架构的特点对传统协同过滤算法进行并行化扩展,设计了基于用户的分布式协同过滤算法;最后通过重定义序列组合的MapReduce模式将多个子任务串联起来,自动地完成顺序化的执行。实验结果表明,该算法在分布式计算环境下具有良好的推荐精度和推荐效率。
With the era of big data's coming,the amount of application data increases sharply,the result of which gives more and more prominence to a personalized recommendation technique.However,traditional recommendation techniques applied to big data are confronted with some problems,such as low recommendation accuracy,long recommendation time and high network traffic.Therefore,the performance of recommendation degrades drastically.To address this issue,a user co-occurrence matrix recommendation strategy was proposed in this paper.The user for the project's predication rating matrix is got by multiplying user similarity matrix and item rating matrix.Candidate items set is generated for each user using user similarity matrix multiply by item similarity matrix.On this basis,traditional collaborative filtering algorithms were parallel expand according to the feature of distributed processing architecture,and a distributed collaborative filtering algorithm was designed.Finally,multi-sub tasks are in series utilizing combination of redefined MapReduce schema to execute automatically.Experimental results show that our approach achieve better prediction accuracy and efficiency in distribute computing environment.
出处
《计算机科学》
CSCD
北大核心
2016年第S2期428-435,共8页
Computer Science
基金
国家自然科学基金项目(60803086)
国家科技支撑计划子课题(2013BAH21B02-01)
北京市自然科学基金项目(4153058
4113076)资助
关键词
协同过滤
推荐系统
分布式计算
大数据
Collaborative filtering
Recommendation systems
Distributed computing
Big data
