基于Mahout框架的协同过滤推荐引擎的研究与实现
发布时间：2018-11-26 09:03
【摘要】：当今互联网应用的快速扩张给广大用户带来了大量的数据信息，很好地满足了用户在信息时代对信息的需求，特别是Web2.0的迅速发展以及移动互联网的崛起，用户自创和分享内容变得越来越容易，用户生成内容随即大量产生。现有的信息检索技术（如搜索引擎）一定程度上解决了激化的矛盾，但是并不能完全满足社会的需求，当用户自身无法给出有效的关键词，又或者用户没有明确需求时，，搜索引擎被动驱动的缺点暴露无遗。这时推荐系统作为信息检索领域新兴技术应运而生。它依靠其智能挖掘用户需求、主动推送精准信息，很快得到了研究者和市场的关注。 本文的项目目标在于探索构建一个基于Hadoop平台的协同过滤推荐引擎，利用开源框架Mahout实现传统协同过滤算法到MapReduce编程模型的移植。本文首先介绍了推荐引擎的研究背景、选题意义、国内外研究现状，阐述了推荐引擎的理论知识和协同过滤算法；其次，详细叙述了本推荐引擎的总体架构和推荐引擎的算法设计，随后，重点阐述了本推荐引擎的具体实现过程，最后给出了推荐引擎的实验结果及在现实中的初步应用。 本论文的主要贡献包括： 1）设计并实现了基于Hadoop的协同过滤推荐引擎，实现了传统协同过滤算法从单机到MapReduce框架下的移植。 2）设计并实现了Web管理系统管理Hadoop平台上的推荐引擎，支持多策略多任务执行推荐作业。
[Abstract]:Nowadays, the rapid expansion of Internet applications has brought a lot of data information to the vast number of users, which has well met the information needs of users in the information age, especially the rapid development of Web2.0 and the rise of mobile Internet. It becomes easier and easier for users to create and share content, and user-generated content is generated in large quantities. The existing information retrieval technology (such as search engine) has solved the intensified contradiction to some extent, but can not completely meet the needs of the society, when the user can not give effective keywords, or when the user does not have a clear demand. The shortcomings of passive drive of search engine are exposed. At this time, recommendation system as a new technology in the field of information retrieval came into being. It relies on its intelligent mining of user needs and proactively pushing accurate information, which has attracted the attention of researchers and markets. The goal of this paper is to explore the construction of a collaborative filtering recommendation engine based on Hadoop platform and implement the migration of traditional collaborative filtering algorithm to MapReduce programming model using open source framework Mahout. This paper first introduces the research background of recommendation engine, the significance of choosing the topic, the research status at home and abroad, and expounds the theoretical knowledge of recommendation engine and collaborative filtering algorithm. Secondly, the overall architecture of the recommendation engine and the algorithm design of the recommendation engine are described in detail. Then, the implementation process of the recommendation engine is described in detail. Finally, the experimental results of the recommendation engine and its preliminary application in reality are given. The main contributions of this thesis are as follows: 1) the collaborative filtering recommendation engine based on Hadoop is designed and implemented, and the traditional collaborative filtering algorithm is transplanted from single machine to MapReduce framework. 2) the recommendation engine on Web management Hadoop platform is designed and implemented, which supports multi-strategy and multi-task recommendation.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3
本文编号：2358076
[Abstract]:Nowadays, the rapid expansion of Internet applications has brought a lot of data information to the vast number of users, which has well met the information needs of users in the information age, especially the rapid development of Web2.0 and the rise of mobile Internet. It becomes easier and easier for users to create and share content, and user-generated content is generated in large quantities. The existing information retrieval technology (such as search engine) has solved the intensified contradiction to some extent, but can not completely meet the needs of the society, when the user can not give effective keywords, or when the user does not have a clear demand. The shortcomings of passive drive of search engine are exposed. At this time, recommendation system as a new technology in the field of information retrieval came into being. It relies on its intelligent mining of user needs and proactively pushing accurate information, which has attracted the attention of researchers and markets. The goal of this paper is to explore the construction of a collaborative filtering recommendation engine based on Hadoop platform and implement the migration of traditional collaborative filtering algorithm to MapReduce programming model using open source framework Mahout. This paper first introduces the research background of recommendation engine, the significance of choosing the topic, the research status at home and abroad, and expounds the theoretical knowledge of recommendation engine and collaborative filtering algorithm. Secondly, the overall architecture of the recommendation engine and the algorithm design of the recommendation engine are described in detail. Then, the implementation process of the recommendation engine is described in detail. Finally, the experimental results of the recommendation engine and its preliminary application in reality are given. The main contributions of this thesis are as follows: 1) the collaborative filtering recommendation engine based on Hadoop is designed and implemented, and the traditional collaborative filtering algorithm is transplanted from single machine to MapReduce framework. 2) the recommendation engine on Web management Hadoop platform is designed and implemented, which supports multi-strategy and multi-task recommendation.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.3
【参考文献】
相关期刊论文 前1条
1 许海玲;吴潇;李晓东;阎保平;;互联网推荐系统比较研究[J];软件学报;2009年02期
本文编号：2358076
本文链接：https://www.wllwen.com/kejilunwen/sousuoyinqinglunwen/2358076.html
