新型Elman混沌神经网络的流量预测

Traffic Prediction of New Chaotic Elman Neural Network

根据实际网络中测量得到的网络流量数据,提出一种改进型Elman神经网络模型——季节性输入多层反馈Elman网络。在网络权值的训练过程中引入混沌搜索机制,利用Tent映射的遍历性进行混沌变量的优化搜索,以减少数据冗余,解决局部收敛问题。实验结果表明,该模型及其算法有效提高了网络的训练速度及网络流量的预测精度。

According to a large amount of network traffic data collected from the actual network, this paper proposes a new modified Elman neural network named Season＇,d Input Muhilayer Feedback Elman（SIMF Elman）. Chaos searching is introduced into model training and uses the ergodicity of the Tent map to search the chaotic variables. Thus the data redundancy is reduced and local optimum problem is solved. Experimental results show that new model and strategy can improve the network training speed and forecast accuracy of network traffic.