Research and Application of Genetic Algorithm Based Optimization of BP Neural Network Combinatorial Model

In this paper, we use multiple linear regression model and genetic algorithm based optimized BP neural network algorithm (GA-BP) for prediction and application study comparison, using seven pre-processed feature vectors as input. The results show that the SAMPE and R2 of multiple regression are 24.25% and 0.79, respectively, and the SMAPE and R
²
of GA-BP model are 4.7851% and 0.82, respectively; the prediction accuracy (measured using SMAPE) of GA-BP model for each variable is concentrated around 10%, which indicates the better fit and robustness of the model, reflecting the genetic algorithm optimized BP neural network model, with a view to providing some reference significance and value for prediction in other fields.