【作者】 陈浩；

【导师】 过弋；

【作者基本信息】 华东理工大学 ， 计算机软件与理论， 2014， 硕士

【摘要】 随着互联网和移动设备的发展,人与人之间的互动和联系越来越依赖社交网络。微博作为最流行的社交网络平台之一,在信息和言论传播、用户沟通交流等方面所扮演的角色越来越重要。微博影响力作为用户的重要衡量指标,是微博关系的基础,用户影响力越大,所受到的关注程度也就越高,对网络的影响、对信息的传播作用也就越大。通过影响力大的用户进行信息扩散、舆论导向、商品推介和宣传,无疑将收到事半功倍的效果,这对于社会信息传播和商业营销来说意义重大。在当下IT领域,不管是学术界还是业界,不管是企业、媒体还是技术人员,似乎都在谈论“大数据”。从技术的角度来说,Hadoop是大数据最重要的标签之一。Hadoop是一个能够让用户简易架构和使用的分布式计算平台,用户可以便捷地在Hadoop上开发运行处理海量数据的应用程序。本文首先研究与讨论了Hadoop平台及其相关技术,以及传统微博用户影响力评定方法,比如追随者数量排名算法、PageRank排名算法和用户行为权值排名算法等。在此基础上提出了基于PageRank改进的用户影响力排名算法UserRank。UserRank算法从用户自身质量及其追随者(即follower)质量入手,考虑追随者数量、追随者质量、评论率、转发率和是否微博认证用户等因素,全面分析得到用户微博影响力指数。最后,UserRank算法在搭建的Hadoop集群上实现,实验结果表明UserRank算法相对于追随者数量排名算法、PageRank算法等都具有很大优势,信息更充分,排名更真实,客观地反映出用户的实际影响力。更多还原

【Abstract】 Nowadays, the social networking platforms are becoming so indispensable to people. Microblog, one of the most popular social networking platforms, plays an increasingly important role on information spreading and user communication. The user influence ranking is one of the most important indexes of user, and it is the basis of user relationship. The greater the user influence, the greater the effect on information spreading. By analyzing the user basic information data and user behavior data with data mining, we can get the user influence ranking. It can not only provide technical support and solutions for the microblogging platform, but also make a profit in cooperation with advertisement owners.In the IT field, enterprises, media and technical personnel, are all talking about "big data". From the point of technology, Hadoop is one of the most important symbols of big data. Hadoop is a distributed computing platform which users can easily set up and use. Users can develop and run big data process program on Hadoop.This dissertation discusses the Hadoop platform and its related technology, and investigates the classical microblog user influence assessment algorithms, such as followers ranking and PageRank. Based on the classical microblog user ranking methods, this dissertation proposes an improved user ranking algorithm named UserRank. The UserRank algorithm is based on the classical PageRank and MapReduce programming model, which processes massive datasets with a cluster of machines on Hadoop platform. The UserRank takes the number and the quality of followers, the forwarding rate, the comment rate and whether the user is verified into consideration, and it uses comprehensive analysis to get the microblog user ranking. The UserRank algorithm runs on the Hadoop platform, and the experiment and evaluation result testifies that the UserRank algorithm can achieve a more accurate and effective ranking result.更多还原