一种利用自组织映射和径向基函数神经网络的网络拥塞预测方法

A Mathod of Network Blocking Forecasting about Kohonen Self-Organizing Maps and Radial Basis Function Network

文中提出了一种利用自组织映射(KSOM)和径向基函数(KR)神经网络进行网络拥塞预测的方法.目前的研究表明,预测网络拥塞还存在一些问题,尤其在数据集比较小的时候.因此,为了使网络拥塞问题预测精度高,在预测过程中有必要考虑原有的数据集中每个数据之间的关系.现在为了获得更多的有价值的位置信息,采取了一系列的措施去满足不同数据的情况,包括使用自组织映射神经网络和径向基函数神经网络算法.这一过程使网络能满足不同类型的数据.在本文网络拥塞预测中,采用同一原始数据集,分别对利用自组织映射和径向基函数神经网络的算法和另外两种算法的性能进行比较.实验结果表明,利用自组织映射和径向基函数神经网络的算法具有更好的效果.

We propose an adaptive Kohonen Self-Organizing Maps and Radial Basis Function Network-based method （KR） for network blocking forecasting in the paper. It shows that there are some problems in the network blocking forecasting now, especially when the data set is just small. Therefore, for achieving high accuracy in the network blocking forecasting, it is necessary to consider the relationships between each data within the original data set in the forecasting process. Now to get more valuable position information, a series of processes including Kohonen neural network and RBF network is proposed to meet the types of different data. The process makes the network can meet the different kinds of data. In this application to a city＇s network blocking forecasting, we investigate KR＇s and two other algorithms performance on a original data set. The comparison of experimental results shows that KR is better location performance than others.