基于用户的协同过滤推荐算法MapReduce并行化实现

The MapReduce Parallelization of User-based Collaborative Filtering Recommendation Algorithm

基于用户的协同过滤推荐算法是应用范围广且应用效果较好的推荐算法之一。传统单机模式下运行的基于用户的协同过滤推荐算法在面对海量数据时存在严重的性能瓶颈问题,很难满足实际计算需求,而基于MapReduce的并行计算框架为解决该问题提供了新思路。MapReduce是Hadoop开源框架的核心计算编程模型,MapReduce的设计目标是方便编程人员在不熟悉分布式并行编程的情况下,可将自己的程序运行在分布式系统上。根据基于用户的协同过滤推荐算法特点,提出MapReduce并行化实现方法。实验结果表明,在MapReduce并行计算框架下实现的基于用户的协同过滤推荐算法在算法性能及稳定性方面都取得了理想效果。

The recommended algorithm based on collaborative filtering of users is a recommended algorithm which has a wide range of applications and is effective in practical applications. However, the traditional recommendation algorithm based on us er based collaborative filtering running in stand alone mode encounters a serious performance bottleneck in the case of massive data and is difficult to meet the actual computing requirements. The MapReduce based parallel computing framework provides a new solution to this problem. MapReduce is a kernel computing programming model of Hadoop open source framework. Ma pReduce is designed to facilitate programmers to run their own programs on distributed systems without being familiar with dis tributed parallel programming. Based on the research of the characteristics of user based collaborative filtering recommendation algorithm, this paper proposes a method based on MapReduce parallel computing framework. The experimental results show that the proposed user based collaborative filtering algorithm based on MapReduce parallel computing framework achieves the desired performance in terms of performance and stability of the algorithm.