基于粒子群小波神经网络的网络流量预测模型研究

发布时间：2018-04-06 02:13

本文选题：网络流量预测　切入点：小波变换　出处：《西安电子科技大学》2014年硕士论文

【摘要】：随着互联网规模的不断扩大和多样化网络业务的不断出现，网络流量数据呈现出越来越错综复杂的行为特征，如何对网络进行有效管理并使得网络提供更好的服务质量成为人们越来越关心的问题。其中，如何建立一个有效而准确的预测模型对网络流量进行预测，成为一个具有一定挑战性的研究热点，它对网络多样化的性能评价、拥塞控制、大规模网络规划设计以及业务的服务保障等重要问题的研究，都具有十分重要的意义。论文对网络流量预测模型进行了研究。论文首先对网络流量的特征以及网络流量预测模型的研究现状进行了总结。然后，介绍了由小波变换和神经网络构成的小波神经网络预测模型(WNN模型)，并重点介绍了我们以前工作中提出过的一种使用遗传算法优化小波神经网络的预测模型(WGANN模型)。接着，论文提出了一种基于粒子群小波神经网络的网络流量预测模型(WPSONN模型)，为了提高预测精度并优化网络收敛速度，我们将比遗传算法收敛更快并具有更高预测精度的全局搜索优化的粒子群算法引入预测模型，对BP神经网络的权值及阈值进行优化。该模型使用具有多分辨率和单支重构特点的小波变换，将网络流量训练样本和预测样本分别分解为低频流量和高频流量，分别用于训练和预测。在使用训练样本对神经网络训练（即网络学习）的过程中，使用粒子群算法，通过重复迭代优化BP神经网络各层之间所有的连接权值及阈值，得到性能较优的神经网络。在预测网络流量时，将预测样本的低频和高频流量数据输入训练好的神经网络，分别得到各预测结果，，然后将各分量进行叠加，得到预测模型的预测结果。最后，论文对提出的预测模型，与WGANN预测模型和WNN预测模型进行了对比分析。针对本文所提出的预测模型，论文进行预测实验。实验结果表明，与WGANN预测模型和WNN预测模型相比较，提出的WPSONN预测模型具有更高的预测精度，并具有更快的网络收敛速度，是一种有效的网络流量预测模型。
[Abstract]:With the continuous expansion of the scale of the Internet and the emergence of diversified network services, network traffic data show more and more complex behavior characteristics.How to effectively manage the network and make the network to provide better quality of service has become a growing concern.Among them, how to establish an effective and accurate prediction model to predict network traffic has become a challenging research hotspot.The research of large-scale network planning and service security is of great significance.The network traffic prediction model is studied in this paper.Firstly, the characteristics of network traffic and the research status of network traffic prediction model are summarized.Then, the prediction model of wavelet neural network, which is composed of wavelet transform and neural network, is introduced, and a prediction model using genetic algorithm to optimize wavelet neural network is introduced in detail.Then, a network traffic prediction model based on particle swarm optimization wavelet neural network (PSO) is proposed. In order to improve the prediction accuracy and optimize the convergence speed of the network, a WPSONN model is proposed.The particle swarm optimization algorithm, which converges faster than genetic algorithm and has higher prediction precision, is introduced into the prediction model to optimize the weights and thresholds of BP neural network.The model uses wavelet transform with multi-resolution and single-branch reconstruction to decompose network traffic training samples and prediction samples into low-frequency and high-frequency traffic respectively for training and prediction.In the process of using training samples to train neural networks (that is, network learning), particle swarm optimization (PSO) algorithm is used to optimize all the connection weights and thresholds between layers of BP neural networks through repeated iterations, and a neural network with better performance is obtained.When forecasting network traffic, the low-frequency and high-frequency flow data of the predicted samples are input into the trained neural network, and the prediction results are obtained respectively, and then each component is superposed to obtain the prediction results of the prediction model.Finally, the proposed prediction model is compared with the WGANN prediction model and the WNN prediction model.According to the prediction model proposed in this paper, the prediction experiment is carried out in this paper.The experimental results show that compared with the WGANN and WNN prediction models, the proposed WPSONN prediction model has higher prediction accuracy and faster network convergence speed, so it is an effective network traffic prediction model.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.06;TP18

【参考文献】