基于大数据的IPTV故障定位系统的设计与实现

发布时间：2018-03-13 20:01

本文选题：IPTV　切入点：大数据　出处：《南京邮电大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来IPTV用户的数量在快速增长,IPTV网络的系统结构日趋复杂,故障发生的原因也日趋多样化、综合化。传统的故障定位方法主要依靠运维人员的经验及部署网络探针,但是在大用户量的情况下难以处理迅速定位IPTV故障。IPTV的服务质量是影响用户满意度的重要因素,因此运营商迫切的需要一种能够快速定位IPTV故障的新技术、新方法,从而提高服务质量,提高用户满意度。本论文分析并设计实现了一种大数据环境下的IPTV故障定位系统。首先,本论文对IPTV用户的KPI数据和报障数据进行了清洗、分析。由于KPI数据中包含较多互相关联的指标,我们在做了一系列统计分析的基础上,采用相关性算法与聚类算法相结合的方式筛选出了用户KPI数据中的代表性指标。在代表性指标筛选完成之后,基于机器学习理论,选取合适的模型建立了质差用户筛选模型,得到IPTV用户中的观看体验较差的用户。具体而言,在该部分中,针对所处理的数据特点,对AdaBoost算法进行了多方面的改进,包括样本初始权重的赋值方式以及样本权重的更新方式,以提高算法预测的准确率。再根据质差用户以及用户-设备关联表建立故障设备筛选模型,实现IPTV故障定位。最后通过仿真实验验证了本论文所提出的设计方案,通过与其他模型的对比,验证了改进的AdaBoost算法在提高IPTV系统故障定位准确率方法的效果。此外,本论文对大数据环境下IPTV故障定位系统进行了实现,为了使系统具有可靠性、可扩展性、处理速度快等特点,我们在构建系统时以Hadoop平台为基础,实现了数据的分布式存储、分布式计算,并且采用了Ambari对系统进行管理,采用Scribe收集系统日志统一处理,确保系统能够稳定、高效的运行。
[Abstract]:In recent years, the number of IPTV users is increasing rapidly. The system structure of IPTV network is becoming more and more complex, and the causes of the faults are becoming more and more diversified and integrated. The traditional fault location methods mainly rely on the experience of operators and the deployment of network probes. However, in the case of large number of users, it is difficult to deal with the rapid location of IPTV fault. The quality of service (QoS) is an important factor affecting the customer satisfaction. Therefore, operators urgently need a new technology and method to quickly locate the IPTV fault. In order to improve the quality of service and improve customer satisfaction, this paper analyzes and designs a IPTV fault location system under big data environment. Firstly, the KPI data and obstacle data of IPTV users are cleaned in this paper. Analysis. Because the KPI data contain more interrelated indicators, we have done a series of statistical analysis, The representative indexes in user KPI data are screened by combining correlation algorithm with clustering algorithm. After the representative index selection is completed, based on the theory of machine learning, the quality difference user screening model is established based on the machine learning theory. In particular, in this part, according to the characteristics of the data processed, the AdaBoost algorithm is improved in many aspects. It includes the assignment method of sample initial weight and the updating method of sample weight to improve the accuracy of algorithm prediction. Then the fault equipment screening model is established according to the quality difference user and user-device association table. Finally, the design scheme proposed in this paper is verified by simulation experiments. Compared with other models, the improved AdaBoost algorithm is proved to be effective in improving the accuracy of fault location in IPTV system. In this paper, the IPTV fault location system under big data environment is implemented. In order to make the system reliable, extensible and fast, we build the system on the basis of Hadoop platform and realize the distributed storage of data. Distributed computing, and the use of Ambari to manage the system, Scribe collection system log unified processing to ensure that the system can run stably and efficiently.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN949.292

【参考文献】