Optimized Fault Diagnosis Algorithm under GAN and CNN Hybrid Model

Abstract

The network fault diagnosis algorithm is one of the most important algorithms to solve network transmission problems. In traditional network fault diagnosis, it mainly analyzes and troubleshoots the fault manually by comparing the alarm information of the network performance index with the expert experience database. Diagnosis methods based on manpower analysis will take up much material resources and manpower and increase maintenance costs. Therefore, there is an urgent need for a more efficient and intelligent fault diagnosis technology. In addition, as the current network is becoming more and more complex, and based on this consideration, a semi-supervised fault diagnosis algorithm is studied in this paper. It is a combination of the GAN and CNN models. Meanwhile, the method of combining relief and mutual information is applied to reducing the dimensionality of network feature parameters, and the optimal feature combination is selected. The fluctuations in the convergence of the generated confrontation network model are stabilized. Moreover, the simulation software is used to build the heterogeneous wireless network scenario studied. Meanwhile, an improved fault diagnosis model is constructed to verify that in the case of both GAN and CNN models, the accuracy of fault diagnosis algorithm can reach 98.6%, which is significantly higher than other comparative analysis methods. It has contributed to ensuring the user’s service experience and reducing the cost of network maintenance and operation.