Enhancing corn industry sustainability through deep learning hybrid models for price volatility forecasting

Area of Science:

• Agricultural Economics
• Time Series Forecasting
• Machine Learning

Background:

• Corn price volatility creates market uncertainty, impacting farmers' decisions and hindering sustainable agricultural investments.
• This instability threatens the long-term viability and sustainability of the corn sector.
• Accurate short-term price forecasting is crucial for mitigating these challenges.

Purpose of the Study:

• To propose and evaluate a novel multi-module wavelet transform-based fusion forecasting model for short-term corn price volatility.
• To enhance the accuracy and robustness of maize price predictions.
• To contribute to the sustainability of the corn industry by reducing price uncertainty.

Main Methods:

• Development of the TLDCF-TSD-BBF model, integrating three-layer decomposition combined dual-filter time-series denoising (TLDCF-TSD), bidirectional time-convolutional enhancement network (BiTCEN), bidirectional long- and short-term memory network (BiLSTM), and frequency-enhanced channel attention mechanism (FECAM).
• TLDCF-TSD decomposes price series for feature extraction and noise reduction.
• BiTCEN and BiLSTM capture short- and long-term dependencies, respectively, while FECAM refines feature focus.

Main Results:

• The proposed TLDCF-TSD-BBF model demonstrated superior performance compared to baseline models across multiple evaluation metrics.
• Specific performance indicators include low Mean Absolute Error (MAE) and Mean Squared Error (MSE), and high R-squared (R2) values.
• The model achieved MAE values ranging from 0.0055 to 0.0137, MSE from 0.0001 to 0.0002, MAPE from 0.8456 to 1.7567, and R2 from 0.9888 to 0.9955 across different datasets.

Conclusions:

• The TLDCF-TSD-BBF model effectively predicts short-term corn price volatility, offering a valuable tool for market participants.
• The integrated approach of wavelet decomposition, convolutional networks, LSTMs, and attention mechanisms enhances forecasting accuracy and robustness.
• The study validates the model's efficacy using Chinese corn price data, supporting its potential to improve agricultural sustainability.