Optimization of Neural Network Algorithm and Its Application Based on Particle Swarm

In view of the fact that when BP neural network algorithm is trapped into local extremum and the extremum converges to local minimum point, the convergence rate becomes slow, neural network structures are different, and there is a contradiction between application examples and network scale, this paper uses particle swarm optimization algorithm to optimize initial weights and thresholds of BP neural network. This method effectively enhances the ability of BP algorithm to handle nonlinear problems, and improves BP algorithm's convergence speed and ability to search global optimal values at the same time. A project case was selected for empirical analysis. The empirical results show that the new algorithm model has a significant improvement in algorithm efficiency, accuracy and other aspects, and it is of high practical value.