园区网的网络故障状态预测方法

Network faults prediction method for park network

为及时处理园区网可能发生的网络故障,提出了一种网络故障状态预测方法。首先利用采集的园区网网络运行数据样本做数据挖掘,获得网络运行参数的ARMA预测函数及网络故障预测的BP神经网络模型;做网络故障状态实时预测时,在园区网的关键监测点中利用SNMP协议采集网络运行参数,通过这些参数预测下面几个时段的网络运行参数,进而利用BP网络预测网络监测点的故障状态。实验结果表明,网络运行状态预测的准确率能够达到80%以上,可实现基本的网络故障状态预测。

To deal with the possible network failures in the park network instantly,a newmethod of network faults prediction was proposed. Firstly,the ARMA prediction function to make data mining on the data sample was used. Then the predict function of network operation parameters,and BP neural network model for network failure prediction was obtained. The collection network operating parameters of SNMP protocol on the critical monitoring point of the campus network was applied while the real-time network faults was predicted. Therefore several hours in advance by the above parameters was predicted whenever the network was running. Furthermore the Back Propagation to predict the net failures was used. The experimental results showthat the operation accuracy rate reaches at more than 80% of the network state prediction by using the proposed the method.