云计算环境下的电子商务推荐系统关键技术研究

发布时间：2018-01-05 09:48

本文关键词：云计算环境下的电子商务推荐系统关键技术研究　出处：《北京信息科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：大数据时代下,互联网数据量呈爆炸式增长,如何在海量数据信息中快速获取到有价值的信息,已经成为一项极具挑战的工作。在此背景下,电子商务推荐系统应运而生,它可以挖掘出用户的不同行为偏好,为不同用户提供个性化的推荐服务。目前推荐技术面临着诸多问题,如数据稀疏问题、可扩展性问题、推荐实时性问题、冷启动问题等。云计算是基于互联网的相关服务的增加、使用和交付模式,拥有强大的计算能力和存储能力。因此本文是利用云计算技术对电子商务推荐系统关键技术进行研究改进。论文的主要研究内容有以下三部分:(1)云计算技术分析与研究。对Hadoop集群运作原理,以及Hadoop的主要实现技术包括HDFS分布式文件系统、MapReduce编程模型和HBase数据库等技术进行分析与研究,明确了云平台下的推荐算法设计思路和实现方法。(2)算法研究。在基于云计算技术进行推荐算法改进的设计思路的基础上对算法进行优化改进,具体的实现步骤是利用MapReduce编程模型对算法进行分解合并操作,把算法分成必须串行计算部分和可以并行计算部分,最大化的实现算法的并行化计算,与此对应若干个MapReduce作业过程,每个作业过程包括一个Map函数的分解过程和一个Reduce函数的合并过程。(3)实证研究。以Netflix数据集为数据实证研究对象,在虚拟机上搭建Hadoop集群平台,对本文改进后的算法进行实证研究分析,选取算法的运行时间、加速比、平均绝对误差等测试指标对算法的计算能力进行测试分析总结实验结果。
[Abstract]:Under the era of big data, the Internet data shows explosive growth, how in the massive data rapid access to valuable information, has become a very challenging work. In this context, e-commerce recommendation system came into being, it can not dig out the peer user's preferences, to provide personalized recommendation service for different the user recommended at present technology faces many problems, such as data sparsity, scalability, real-time recommendation, cold start problem. Cloud computing is to increase the related services based on the Internet, the use and delivery mode, has powerful computing ability and storage ability. So this is the use of cloud computing technology for electronic the key technology of Commerce recommendation system is studied. The main contents of this paper are the following three parts: (1) cloud computing technology analysis and research group. Operation principle of Hadoop set in Including the HDFS distributed file system to achieve the main technology and Hadoop, analysis and Research on MapReduce programming model and HBase database technology, the recommendation algorithm design ideas and realization method of cloud platform. (2) algorithm. Based on cloud computing technology based on the design idea of algorithm improvement on the algorithm optimization improvement of the specific implementation steps of algorithm is to decompose the merge operation using the MapReduce programming model, the algorithm must be divided into serial computation part and can realize parallel computing, algorithm and maximization of calculation, and this corresponds to a number of MapReduce operations, each operation process including the merger process decomposition process of a Map function and a Reduce function. (3) empirical research. Based on Netflix data sets as the research object of empirical data, set up in the virtual machine cluster on Hadoop platform In this paper, the improved algorithm is empirically researched and analyzed, and the algorithm's running time, speedup ratio, average absolute error and other test indexes are selected to test and analyze the computing power of the algorithm, and the experimental results are summarized.

【学位授予单位】：北京信息科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3;F724.6

【参考文献】