The Design of a Compound Neural Network-Based Economic Management Model for Advancing the Digital Economy

Abstract

The rapid progress of the digital economy has brought forth a myriad of complexities in economic governance, particularly in the domains of stocks and network finance. The authors propose the exploration of an innovative economic management model founded on the compound neural network framework. Central to this approach is the utilization of the deep bidirectional long and short-term memory neural network model (Bi-LSTM) as the primary instrument for predictive analysis, complemented by the refinement and enhancement provided by the Markov chain model. Through comparative analysis of experiments, it is found that although the forecast price of this model has a certain lag, it has a more accurate judgment than other prediction models, and the accuracy and recall rate reach 87.66% and 86.31%. At the same time, the error evaluation index R2 is very close to the upper limit 1 of the index, and the mean absolute error MAE Hill inequality coefficient; TIC root; mean square error; RMSE; and symmetric mean percentage error (SMAPE) are 0.2654, 0.0124, 0.3481, and 0.3531, respectively.