Research on Combined Prediction Model based on BP Neural Network Model and Multiple Regression Analysis Model

This paper proposes a combined prediction model based on BP neural network and multiple regression analysis, aiming to improve the accuracy of prediction tasks in complex systems by integrating the advantages of the two models. The BP neural network is capable of extracting potentially complex patterns from a large amount of data by virtue of its powerful nonlinear fitting ability, while the multiple regression analysis on the other hand, performs well in dealing with linear relationships between multiple independent variables and dependent variable. In the process of model construction, the importance of different factors was first assessed by the random forest algorithm, and the key factors were screened out; then the correlation between each factor and the prediction target was analysed by using the Spearman test and the independent sample t-test. On this basis, a BP neural network is combined to perform nonlinear prediction, and a multiple linear regression model is used to model the relationship between multiple factors and target variables. The experimental results show that the combined model can effectively improve the prediction accuracy with good adaptability and robustness. However, the model suffers from high computational complexity when dealing with large-scale data and is sensitive to hyperparameter selection. Future research will further improve the predictive ability and applicability by optimising the data processing methods, improving architecture and introducing more external factors.