Research On Fault Diagnosis Methods Of 5G Networks Based On Artificial Intelligence

With the rapid development of mobile communication technologies,although the proposal and application of heterogeneous networks have met different business requirements,it also increases the complexity of mobile communication networks.At the same time,with the large-scale commercial use of 5G networks,the mobile communication networks will become more complex and heterogeneous due to the introduction of enhanced new technologies.The complexity of network operation and maintenance forces mobile operators to find new strategies to remain competitive.However,most of the existing network fault diagnosis methods rely on manual testing and time accumulation,which have the problems of long optimization cycle and high resource consumption.Therefore,from the perspective of big data and artificial intelligence,this thesis has made an in-depth study on the intelligent network fault diagnosis methods.Based on the introduction of existing network fault diagnosis methods,this thesis mainly carries out the research work on the subject from the following three aspects:(1)Aiming at the problem of how to perform 4G/5G networks fault diagnosis accurately when there are few effective labeled samples,a cellular network fault diagnosis algorithm based on Graph Convolutional neural Network(GCN)is proposed.First,the common failure types of 4G/5G networks are analyzed.After that,the e Xtreme Gradient Boosting(XGBoost)algorithm is used to select the optimal subset of network parameters in the collected network parameter dataset,and the graph structure is constructed with data in the dataset as nodes and similarities between nodes as edges.Finally,GCN is used to extract features from the graph data,complete the classification task for nodes,and predict the fault types of networks.A large number of experiments are carried out based on the simulation dataset,results show that compared with a variety of existing algorithms,the proposed method can effectively improve the performance of network fault diagnosis with a small number of labeled samples.(2)Aiming at the shortcomings of GCN based cellular network fault diagnosis algorithm in practical application,on the basis of using the network parameter dataset collected in the real network scene,first,the real dataset is expanded by using Generative Adversarial Network(GAN)to solve the problem of too few labeled samples and unbalance distribution of sample categories.In the process of network fault diagnosis,the network parameter dataset is pre diagnosed by using Naive Bayesian Model(NBM)combined with expert knowledge,and the topological association diagram between the data is constructed according to the results of pre-diagnosis.After that,the association diagram as the prediagnosis prior knowledge and the training dataset are input into the GCN model together for model training,and the original GCN is improved to adjust the influence of the pre-diagnosis prior knowledge and the size of the training dataset on the model accuracy in the process of model training.The experimental results show that the data generated by GAN conforms to the distribution of the real data.At the same time,because the proposed algorithm combines pre-diagnosis knowledge and deep learning effectively,it can achieve higher diagnosis accuracy compared with other traditional algorithms.(3)Designed and completed the visualization platform of network fault diagnosis based on big data.The network fault diagnosis visualization platform is mainly realized through front-end and back-end technologies.The platform can analyze and process the network parameter dataset through the back-end service,and then use the front-end technologies to visually display the network fault diagnosis results.