Research On Several Algorithms Of Faults Diagnosis Of Network Based On Deep Learning

With the era of big data and the rise of mobile Internet,heterogeneous wireless networks have undergone tremendous changes.The features such as complexity,heterogeneity,dynamism and openness slowly appear,and the quality of service of users and various services of Service(QoS)requirements are becoming higher.Therefore,how to realize the fault diagnosis,analysis and optimization of complex heterogeneous wireless networks is of great research value and application value.However,the traditional methods of network fault diagnosis,such as optimization of existing networks,fault monitoring and fault diagnosis,bring about huge operating costs and inaccurate model establishment.Therefore,from the perspective of big data and machine learning,the paper deeply studies the fault diagnosis methods of complex heterogeneous wireless networks.Based on the study of existing methods of network fault diagnosis,the following three aspects are introduced:(1)A network fault diagnosis method based on long and short memory neural network is proposed.This method studies the complex mapping relationship between physical network and virtual network after network virtualization.Based on the characteristics of physical nodes and physical links,this paper analyzes the important network characteristic parameters in the network and makes use of the long-short-time memory neural network extract,map fault features to higherdimension space,and enhance the identifiability of fault categories.Simulation shows that this method has higher fault recognition rate.(2)A two-phase fault diagnosis algorithm based on convolutional neural network is proposed.Based on the analysis of the causes of heterogeneous wireless network failures and the fact that the wireless network resources are precious,the method of features selection is used to select the network parameters that have a great influence on the nodes.The characteristics of the timing distribution of the network parameters are monitored.Based on the monitoring results,convolutional neural network classify the network operatiing status.The simulation results show that the method has better performance in controlling resource consumption by nodes and has higher accuracy in network fault diagnosis.(3)OPNET-based network fault scenarios in heterogeneous network environment is designed and implemented.Based on the simulation scenario of OPNET network faults,the fault data in the network is obtained.The model train the network fault data to classify the network abnormal data.This part introduces the setting process of OPNET fault scene,the process of data preprocessing,the process of feature selection,the setting of detailed parameters of convolutional neural network and the setting of training process.