Early Detection of Network Fault Using Improved Gray Wolf Optimization and Wavelet Neural Network

To address the problem of diagnostic accuracy and stability degradation caused by random selection initial parameters for wavelet neural network (WNN) fault diagnosis model, this paper proposes a troubleshooting model based on improved gray wolf algorithm (IGWO) network. First, convergence factor policy weight update are redesigned in IGWO algorithm. This study uses nonlinear to balance global local search capabilities dynamically adjusts weights according adaptability head α strengthen its leadership position. Thereafter, biases WNN optimized using During backpropagation error, momentum factors introduced prevent from falling into optimization. Experimental results show that is far better than GWO terms speed accuracy. Furthermore, average IGWO-WNN KDD-CUP99 dataset reaches 99.22%, which 1.15% higher significantly improved.