基于改进时间卷积网络的局域网异常预测仿真

Research on Simulation of LAN Anomaly Prediction Based on Improved Time Convolution Network

深度学习的快速发展时刻影响着各领域的发展,针对传统的局域网异常预警存在训练复杂度高、预测精度低、预测时间长等问题,提出一种基于改进时间卷积网络的局域网异常预测方法。首先介绍了时间卷积神经网络;通过对激活函数进行改进,解决传统神经网络存在神经元坏死问题;其次采用自适应线性对全连接层进行替换解决传统网络存在的过拟合问题。最后,为了验证所提出算法的有效性,采用了不同方法与提出的方法进行对比。仿真结果表明,所提算法在预测精度、预测时间以及模型训练复杂度上有较好的提升。

The rapid development of deep learning has always affected the development of various fields.In view of the problems of high training complexity,low prediction accuracy and long prediction time in traditional LAN anomaly early warning,this paper proposes a LAN anomaly prediction method based on improved time convolution network.Firstly,the time convolution neural network was introduced,and the problem of neuron necrosis in traditional neural network was solved by improving the activation function;Secondly,an adaptive linearity was used to replace the full connection layer to solve the problem of overftting existing in the traditional network;Finally,in order to verify the effectiveness of the algorithm proposed in this paper,different methods were used to compare with the methods proposed in this paper.The simulation results show that the proposed algorithm has a better improvement in the prediction accuracy,prediction time and model training complexity.