基于Hadoop的移动互联网用户移动性预测模型研究

发布时间：2018-04-02 19:51

本文选题：移动互联网　切入点：移动性预测　出处：《北京邮电大学》2017年硕士论文

【摘要】：随着移动终端技术的不断发展,移动互联网用户的数量呈现飞速的增长,越来越多的用户通过移动终端,比如手机、平板电脑等设备,从互联网中获取自己需要的资源。这个过程会产生海量的移动通信数据,通过分析这些移动性数据,我们可以掌握用户的移动规律和活动轨迹,深入理解用户的行为特征和移动模式,帮助我们实现移动互联网用户的移动性预测。移动性预测模型的研究是理解人们日常生活习惯、社交关系、个人喜好等行为的重要课题,可以帮助我们解决如应用推荐、疾病预防、城市规划、人口迁徙、国防建设等各种实际问题,对促进移动通信技术的发展有着积极的影响和重大的研究意义。移动通信的过程中,用户产生的数据是海量的,传统的数据处理技术已经无法满足我们的存储和分析要求。所以,本文通过Hadoop分布式软件平台来实现对用户的移动性数据的分析和预测。本文首先介绍了 Hadoop的架构,及其进行大数据存储和计算的基本流程和工作原理。在基于Hadoop的平台上,构建了移动互联网用户的移动性预测模型。随后,研究了用户轨迹里重要位置的提取方法,并基于Point of Interests(POI)识别率指标对这些算法进行了评估与对比,选取出最理想的方法完成了数据的预处理。接着,本文对移动互联网用户进行了移动性分析,来判断用户位置的可预测性。并在熵算法基础上提出了瞬时熵与转移概率结合的算法,完成了对用户轨迹里的每个访问位置的可预测性的判断。可预测性分析的实验结果为移动性预测提供了有力的依据。最后,本文实现了移动互联网用户的移动性预测,并针对传统算法的缺陷提出了改进预测性能的方法,提高了移动性预测的准确度。
[Abstract]:With the continuous development of mobile terminal technology, the number of mobile Internet users is increasing rapidly. More and more users obtain the resources they need from the Internet through mobile terminals, such as mobile phones, tablets and other devices.This process will produce massive amounts of mobile communication data. By analyzing these mobility data, we can grasp the mobility laws and activity trajectories of users, and deeply understand the behavior characteristics and mobility patterns of users.To help us achieve mobile Internet users mobility prediction.The research of mobility prediction models is an important topic to understand people's daily life habits, social relationships, personal preferences and so on, which can help us solve problems such as application recommendation, disease prevention, urban planning, population migration, etc.National defense construction and other practical problems have a positive impact on the development of mobile communication technology and significant research significance.In the process of mobile communication, the data generated by users is huge, the traditional data processing technology can not meet our storage and analysis requirements.Therefore, this paper uses Hadoop distributed software platform to realize the analysis and prediction of user mobility data.This paper first introduces the architecture of Hadoop, the basic flow and working principle of big data storage and calculation.On the platform of Hadoop, the mobility prediction model of mobile Internet users is constructed.Then, the extraction method of the important position in the user trajectory is studied, and these algorithms are evaluated and compared based on the Point of Interests-Poi recognition rate index, and the most ideal method is selected to complete the data preprocessing.Then, this paper analyzes the mobility of mobile Internet users to determine the predictability of user location.Based on the entropy algorithm, an algorithm combining instantaneous entropy and transition probability is proposed, which can judge the predictability of each access location in the user's trajectory.The experimental results of predictive analysis provide a powerful basis for mobility prediction.Finally, the mobility prediction of mobile Internet users is realized, and a method to improve the performance of the traditional algorithms is proposed to improve the accuracy of mobility prediction.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN929.5;TP393.0

【相似文献】