Prediction of US Cotton Futures Price under Different Models

This paper provides a comprehensive test and comparison of four forecasting models selected to forecast U.S. cotton futures prices. First, the ARIMA model performs well in trend analysis and short-term forecasting, especially when dealing with linear time series data. However, when faced with nonlinear data, such as future price forecasts, the accuracy of the ARIMA model falls short. Second, the SVR and MLP models exhibit relatively high forecasting accuracy when dealing with nonlinear data. Unfortunately, however, despite the seemingly reliable prediction results, they failed to pass rigorous hypothesis testing, which limits their practical applications. The BPNN model, on the other hand, stands out due to its excellent performance, which is able to accurately capture the complex relationships in the data and make accurate predictions accordingly. By comparing the prediction results of different models, especially the excellent performance of the BPNN model in nonlinear data prediction, this study provides new ideas and methods to improve the prediction accuracy. This has an important practical application value for fields that require highly accurate prediction.