Mathematical Modeling Method based on Neural Network and Computer Multi-Dimensional Space

In order to predict the data in the macroeconomic database, basic analysis techniques mainly use time series models or statistical regression models, but their effects on processing nonlinear complex systems are not satisfactory. On the other hand, in fact, the factors that affect the future of the dependent variable are not only related to historical time series data, but also have important relationships with many other simultaneously influencing factors, and these simultaneous influencing factors are usually numerous and complex interactions.. The relationship of influence makes the analysis and prediction of the problem more complicated and tedious. Due to all the above factors, traditional forecasting methods are unsatisfactory. Artificial neural network is a new prediction technology that has emerged in recent years. It has the powerful ability to deal with large and complex nonlinear dynamic systems, and has research and prediction applications. This article proposes a neural network-based forecasting model that combines time series data and statistical cross-sectional data analysis. It not only uses the advantages of neural networks to deal with complex nonlinear systems, but also starts with the prediction principles derived from traditional prediction methods. Improve the prediction of a single time series network, and check historical time series data and cross-sectional factor data at the same time, so that the embedded model reflects the historical trend of dynamic growth and causal mapping. In order to confirm the above point of view, this article collected a lot of real macroeconomic data, and established the annual absolute value model with three different methods. The comparative experiment results show that the model is superior to the other two models, which proves the feasibility and effectiveness of the method to a certain extent, has made a certain degree of progress, and has a certain application value in the field of macroeconomic indicator prediction. The accuracy of the method used in this paper has been improved by 12%.