一种优化的RBF神经网络模型用于网络流量预测

AN OPTIMIZED RADIAL BASIS FUNCTION NEURAL NETWORK MODEL APPLIED TO NETWORK TRAFFIC FORECAST

采用径向基RBF神经网络对网络流量数据的时间序列进行建模与预测。采用传统的学习算法对RBF网络训练时,对网络流量数据容易出现过拟合现象,提出了自适应量子粒子群优化AQPSO算法,用于训练RBF神经网络的基函数中心和宽度,并结合最小二乘法计算网络权值,改善了RBF神经网络的泛化能力。实验结果表明,采用AQPSO算法获得的RBF神经网络模型具有泛化能力强、稳定性良好的特点,在网络流量预测中有一定的实用价值。

The time series of network traffic data is modelled and forecasted based on Radial Basis Function neural network （RBFNN） in this paper. Using classical learning algorithm to train radial basis function （RBF） .neural network, it often shows over fitting phenomenon to network traffic data, which lead to lower precision. Adaptive Quantum-behaved Particle Swarm Optimization （AQPSO） algorithm is proposed in order to improve network＇ s performance. By applying AQPSO algorithm to train the central position and width of the basis function adopted in the RBFNN, and computing the network＇ s weights with least-square method, the generalization ability of the RBFNN is improved. Experimental resuhs with real network traffic data sets show that the obtained network model has not only good generalization properties, but also better stability. It illustrates that RBFNN with AQPSO optimization algorithm has the promising application in network traffic data forecasting.