Genetic algorithm in the field of BP neural network optimization

BP neural network, as a traditional supervised artificial neural network, has been widely used in the field of prediction and classification for dealing with nonlinear data problems, but it is easy to get local minima and overfitting, and poor generalization. To address the above problems, this paper uses the powerful macroscopic search ability of genetic algorithm(GA) to optimize the structure and training process of BP neural network, finds the globally optimal weights and thresholds, and proposes a nonlinear data prediction model based on GA optimized BP neural network. The experimental results display that the prediction results obtained by the BP neural network optimized with GA have higher accuracy compared with the single BP prediction model, and the prediction accuracy is 12% higher on average, and the overfitting problem is also solved, and the mean square error(MSE) is reduced to the order of 10
⁻⁴
, which improves the generalization of the model.