基于小波降噪和改进免疫优化的BP模型网络流量预测

Prediction of Network Flow Traffic of BP Network Model Based on Wavelet De-Noising and Improved Immune Optimizing

为了实现大规模复杂网络的流量预测,并改善传统BP网络预测模型存在的收敛速度慢和容易陷入局部最优的缺陷,提出了一种基于小波降噪和改进人工免疫优化BP神经网络的网络流量模型;首先,描述了网络流量预测的基本原理;然后,采用小波包降噪法对网络流量原始采样序列进行降噪处理,在此基础上定义了采用BP网络进行网络流量预测的算法,在确定了神经网络的结构后,采用训练数据和改进的人工免疫优化算法对BP网络中的参数即权值和阀值进行优化,从而得到最终的BP网络流量预测模型;最后,采用1 800组样本中的1 200组训练数据对网络进行训练后得到最终的BP网络模型,再采用剩余的600组测试数据进行流量预测;实验结果证明结合人工免疫算法和BP网络的网络流量预测模型能实现大规模复杂网络的流量预测,且较传统方法相比,具有收敛速度快、训练时间短和预测精度高的优点。

In order to realize the predication for network flow traffic, and conquering the defects of slow convergence efficiency and get- ting the local optimal solution, a prediction model for network flow traffic based on wavelet de--Noising and improved immune optimizing BP network model was proposed. Firstly, the basic principle of predication for network traffic flow was described, then using the wavelet de-- noising method to deduce the noise in the rude data. Finally, the algorithm for predict the network traffic flow was defined, and after the net- work strueture was fixed, the training data and the improved artificial immune algorithm was used to optimize the parameters in the BP model such as weight and threshold, therefore, the final BP model was obtained. Using the 1800 group data to verify the method in this paper, and using 1200 group data to train the BP network model, and use the remained 600 group data to predict network traffic flow. The experiment shows our method combined artificial immune algorithm and BP network can realize the flow prediction in mass and compound network, com- pared with the other methods, it has the advantage of rapid convergence, short training time and high predication accuracy.