基于优化神经网络的无线网络流量预测方法研究

发布时间：2018-02-09 21:04

本文关键词： 无线网络流量建模预测服务质量神经网络稳定小波变换量子遗传算法　出处：《北京邮电大学》2014年硕士论文　论文类型：学位论文

【摘要】：无线网络在信息化社会中扮演着越来越重要的角色。无线网络能够轻易、有效的进行高速通信,为人们的生活提供便利的同时,也为国家经济、政治、军事带来了新的发展契机。随着越来越多的无线宽带网络接入点部署在生产生活中,无线网络规模日益庞大,环境日趋复杂,网络运营商缺乏有效保证无线网络服务质量(QoS)的手段,对于无线网络流量的模型、特征、可靠性需要进一步研究,使得保障网络QoS、维护网络安全、网络故障检测等工作难以深入开展,网络流量的建模预测已经成为解决这一问题的主要工具。本文针对无线网络流量本身和基于优化人工神经网络的建模预测进行了系统研究。为掌握无线网络流量的预测方法,本文首先对无线网络流量数据进行了研究,通过分析其统计特性、相关特性、自相似性、混沌特性等,并与有线网络流量对比,验证了无线网络流量具有更强的分散性、突发性和混沌特性。接着,本文对时间序列预测方法进行了调研,分析了传统时间序列分析、混沌时间序列分析的方法,进一步研究ARIMA模型和混沌RBF神经网络模型的预测方法,发现这些模型在网络流量预测中存在一定缺陷,需要更精确的模型来预测无线流量。然后,本文重点研究BP神经网络、量子遗传算法和小波变换理论,深入探讨BP神经网络的概念、原理和优缺点,分析神经网络优化的方法,提出一种利用量子遗传算法高效的全局搜索能力来优化神经网络的方法。在此基础之上,结合稳定小波变换,利用BP神经网络良好的鲁棒性和非线性处理能力,提出一种基于优化神经网络的混合无线网络流量预测模型,命名为SWT-QGA-BP模型。最后,仿真实验对无线网络流量进行单步、多步预测,结合预测评估指标,对提出的SWT-QGA-BP模型的预测结果进行评价,对比ARIMA模型和混沌RBF神经网络模型,验证了新模型的自适应性和预测性能优越性。提出的SWT-QGA-BP模型能够更加准确高效的对无线网络流量进行预测,有能力为网络保障QOS、网络资源管理、网络安全维护提供必要的助力。
[Abstract]:Wireless network plays a more and more important role in the information society. Wireless network can easily and effectively carry out high-speed communication, provide convenience for people's life, at the same time, it is also good for national economy and politics. As more and more wireless broadband network access points are deployed in production and daily life, the scale of wireless network is becoming larger and larger, and the environment is becoming more and more complex. Network operators lack of effective means to ensure the quality of service (QoS) of wireless networks. The model, characteristics and reliability of wireless network traffic need further study to ensure network QoS and maintain network security. Network fault detection is difficult to carry out in depth, and modeling and forecasting of network traffic has become the main tool to solve this problem. In this paper, wireless network traffic itself and modeling and prediction based on optimized artificial neural network are systematically studied. In order to master the prediction method of wireless network traffic, this paper first studies the wireless network traffic data, through the analysis of its statistical characteristics, correlation characteristics, self-similarity, chaos characteristics, and so on, and compared with the wired network traffic. It is proved that wireless network traffic has more dispersive, sudden and chaotic characteristics. Then, this paper investigates the prediction methods of time series, analyzes the methods of traditional time series analysis and chaotic time series analysis, and further studies the prediction methods of ARIMA model and chaotic RBF neural network model. It is found that these models have some defects in network traffic prediction, and more accurate models are needed to predict wireless traffic. Then, this paper focuses on BP neural network, quantum genetic algorithm and wavelet transform theory, deeply discusses the concept, principle, advantages and disadvantages of BP neural network, and analyzes the optimization method of neural network. This paper presents a method of optimizing neural network by using the high efficient global search ability of quantum genetic algorithm. On this basis, combining with stable wavelet transform, the good robustness and nonlinear processing ability of BP neural network are utilized. A hybrid wireless network traffic prediction model, named SWT-QGA-BP model, is proposed based on optimized neural network. Finally, the simulation experiments are used to predict the wireless network traffic in single step and multi-step. Combined with the prediction evaluation index, the prediction results of the proposed SWT-QGA-BP model are evaluated, and the comparison between the ARIMA model and the chaotic RBF neural network model is carried out. The proposed SWT-QGA-BP model can predict wireless network traffic more accurately and efficiently, and can provide necessary assistance for QoS, network resource management and network security maintenance.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP18;TP393.06

【参考文献】