基于模糊神经网络的网络业务量预测研究

Study of predicting network traffic using fuzzy neural networks

利用神经网络(NN)的自学习能力以及模糊逻辑的动态性和及时性等特点,将模糊逻辑和 NN 有机地结合起来,构造出了五层模糊神经网络(FNN),并用训练 NN 的相应学习算法-BP 算法来训练网络。本文将 FNN 用于网络自相似业务预测研究中,并与单纯的 NN 算法相比较。仿真结果表明,FNN 能很好地预测复杂网络业务,与传统的 NN 算法相比,不仅收敛速度快,且得到更好的预测效果。本文为复杂网络业务流量预测研究提供了一种有效途径。

A five-layer fuzzy neural network (FNN), which utilized both the linguistic, was addressed human-like reasoning of fuzzy systems and the powerful computing ability of neural networks (NN), for predicting the nonlinear self-similar network traffic. The FNN was trained by the Back-propagation algorithms, which was used to train the NN. The simulation demonstrates that the FNN could well predict this traffic. Compared with the sole NN, FNN was superior to NN. FNN not only could converge quickly, but also the error was lower. It supplies a feasible method for predicting the complex network traffic research.