Research on GDP Forecast Analysis Combining BP Neural Network and ARIMA Model

Area of Science:

• Econometrics
• Computational Finance
• Artificial Intelligence

Background:

• Traditional time series models like ARIMA have limitations in capturing complex nonlinearities in economic data.
• Forecasting accuracy is often impacted by numerous external factors, necessitating multivariate approaches.
• Existing GDP prediction models primarily use weighted combinations, with limited exploration of alternative integration methods.

Purpose of the Study:

• To develop an enhanced GDP forecasting model by integrating the strengths of BP neural networks and ARIMA.
• To introduce and validate an error-correction combination method for time series forecasting.
• To compare the performance of the proposed error-correction model against traditional weighted methods and individual models.

Main Methods:

• Utilizing the Box-Jenkins methodology for ARIMA (Autoregressive Integrated Moving Average) model development.
• Implementing a Backpropagation (BP) neural network for nonlinear residual prediction.
• Developing a novel error-correction combination strategy to integrate BP and ARIMA predictions.
• Determining optimal BP neural network parameters, including hidden layer nodes, based on price characteristics.

Main Results:

• The error-corrected GDP forecast model demonstrated superior performance compared to the weighted GDP forecast model.
• The BP neural network achieved a relative prediction error of approximately 2.5% for monthly prices.
• For daily price prediction, the BP neural network achieved a relative error of 1.5%, outperforming the ARIMA model's 2% relative error.

Conclusions:

• The error-correction combination method is an effective strategy for enhancing time series forecasting accuracy.
• BP neural networks offer a robust alternative for economic forecasting, particularly in capturing nonlinear patterns.
• The integrated BP-ARIMA model with error correction provides a more accurate and reliable approach to GDP prediction.