基于人工智能的图卷积神经网络故障诊断方法研究

Research on the fault diagnosis method of graph convolutional neural network based on artificial intelligence

随着移动互联网的快速发展,网络故障诊断技术已成为一个重要的研究方向。在移动网络故障诊断中,由于故障样本数量有限,传统方法难以准确诊断出故障类型。因此,研究提出了一种融合朴素贝叶斯模型(Naive Bayesian Model,NBM)和图卷积神经网络(Graph Convolutional Networks,GCN)的移动网络故障诊断方法。通过GCN与NBM的融合,故障诊断方法能够提取到更多故障数据,用于对故障的识别和诊断。结果表明,模型方法的故障诊断准确率平均值和故障误检平均值分别为92.18%、9.13%;同时模型方法在网络故障分类识别效率为75.00%,且在故障识别开始时的平均时间开销为11 s。所有结果均优于对比算法,这说明所提出的方法能够有效地识别出移动网络故障类型,并具有较高的准确率和鲁棒性。

With the rapid development of mobile Internet,network fault diagnosis technology has become an important research direction.In mobile network fault diagnosis,due to the limited number of fault samples,it is difficult for the traditional methods to accurately diagnose the fault types.Therefore,the study proposes a mobile network fault diagnosis method that fuses Naive Bayesian Model(NBM)and Graph Convolutional Networks(GCN).Through the fusion of GCN and NBM,the fault diagnosis method is able to extract more fault data for fault identification and diagnosis.The results show that the average fault diagnosis accuracy and the average fault false detection of the model method are 92.18%and 9.13%,respectively;meanwhile,the model method has 75.00%efficiency in network fault classification identification and the average time overhead at the beginning of fault identification is 11 s.All the results are better than the comparison algorithms,which indicates that the proposed method can effectively identify the mobile network fault types and has high accuracy and robustness.