Construction of Macroeconomic Data Analysis Model Based on Improved BP Model

In order to improve the analysis accuracy of China's macroeconomic data, a prediction model based on LSTM+BP combined neural network is proposed. By combining the LSTM long short term neural network that can save the long-time sequence data and automatically match the optimal time interval with the BP neural network that completes the weight allocation and update of the neural network through error back propagation, the learning ability of the model on the historical energy dimension data is improved, and the prediction performance of the model is finally improved. The experimental results show that the on the training set, fitting results of the proposed combined model are consistent with the real curve, and the error is always within 0.7, so the fitting effect is better. It can be seen from the verification on the prediction set that the prediction results of the proposed combined model in the past three years are close to the real results. Compared with single neural network models such as LSTM, BP and WNN, the average error of MPAE error is reduced by 5.09%, 3.8% and 0.04% respectively, which proves that the proposed model has better prediction performance and generalization, and is more suitable for the analysis and research of China's macroeconomic data.