中文摘要：

为提高网络流量预测的精度，针对网络流量的非线性特征提出了一种基于新的误差评价准则——最大相关熵准则（MCC）的网络流量预测方法。该方法使用MCC对Elman神经网络进行训练。该评价准则是基于新的相似度函数——广义相关熵（corrent．ropy）函数的概念建立的，此相似度函数以误差概率密度函数的Parzen窗估计和瑞利熵为基础。同时结合MCC和最小均方误差（MMSE）准则提出了一种混合的评价准则MCC-MMSE。针对网络流量的非线性、非高斯性、突变性等特性，分别以MCC、MCC-MMSE准则进行了Elman神经网络的训练，使用训练好的神经网络进行网络流量预测，仿真结果表明预测结果的精度优于以MMSE为准则的Elman神经网络的预测结果。

英文摘要：

With the nonlinear characteristics of network traffic considered, a new network traffic prediction method based on the maximum correntropy criterion （ MCC）, a new error evaluation criterion, was proposed to improve the preci- sion of traffic network prediction. The method uses the MCC to train Elman neural networks, and this evaluation criterion is based on the new concept of a new similarity function, the generalized correlation entropy （correntropy） function, which takes the Parzen window estimation of the error probability density function and the Rayleigh entro- py as the basis. Simultaneously, a mixed evaluation criterion which combines the MCC and the minimum mean square error （MMSE） criterion was presented. In view of the characteristics of traffic networks such as the nonlin- ear, non Gauss, and mutation, the Elman neural network was trained by the MCC and the mixed criterion, respec- tively, and then a trained neural network was used to predict network traffic. The simulation results show that the accuracy of the prediction is superior to the prediction results of the Elman neural network with the MMSE criterion.