【摘要】 随着计算机技术的发展和互联网的快速普及,如何快速的从海量数据中获取用户想要的信息逐步成为用户关注的焦点之一。个性化推荐系统应运而生,通过获取用户在互联网上的日志信息,分析用户的喜爱偏好,从而为用户推荐其可能感兴趣的信息。然而,随着互联网的发展,互联网上充斥的用户日志信息越来越多,个性化推荐系统面临着存储空间的可扩展性与分析计算的效率等瓶颈问题,单纯依靠提升计算机存储空间和计算性能显然不能从根本上解决问题。本文针对该问题,以基于Hadoop分布式文件存储HDFS和分布式计算框架Map/Reduce的工具为基础,针对本文的应用场景对Mahout实现的推荐算法做了优化,实现了一种基于分布式计算框架Hadoop之上的个性化推荐系统架构,并通过实验证明该方案解决了数据存储的可扩展性和计算性能方面的问题,一定程度上提高了推荐的准确率。更多还原

【Abstract】 As the increase of the network data, how to obtain the information quickly is becoming one of the most important problem for users. Personalized recommendation system is used to analysis the users behavior log and get the users favorite preferences, then help the users get what they want fastly. However, more and more information makes to produce a series of problems, such as the scalability,the computing performance. To solve this problem, this paper proposes a solution that designing and implementing a personalized recommendation system framework based on Hadoop, and optimizing the code of the Mahout to improve the accuracy rate of the recommendation system. The experiments results validate that the recommendation accuracy rate is improved and computing performance is enhanced more than the stand-alone mode.更多还原