A genetic algorithm optimized hybrid model for agricultural price forecasting based on VMD and LSTM network

Area of Science:

• Agricultural Economics
• Data Science
• Time Series Analysis

Background:

• Agricultural commodity prices exhibit complex, nonlinear, and nonstationary patterns, challenging traditional forecasting models.
• Existing prediction methods often struggle to capture these inherent complexities, leading to suboptimal accuracy.
• Accurate price forecasting is crucial for informed decision-making by farmers, traders, and policymakers.

Purpose of the Study:

• To develop and evaluate a novel hybrid Variational Mode Decomposition-Long Short-Term Memory (VMD-LSTM) model for enhanced agricultural commodity price prediction.
• To address the limitations of existing models in capturing nonlinear and nonstationary price dynamics.
• To improve the accuracy and reliability of agricultural price forecasts.

Main Methods:

• A hybrid VMD-LSTM model integrating genetic algorithm (GA) optimization for both VMD and LSTM components.
• GA-optimized VMD decomposes price series into sparse intrinsic mode functions (IMFs) for efficient modeling.
• Individual IMFs are forecasted using GA-optimized LSTM, with final predictions ensembled.

Main Results:

• The VMD-LSTM model demonstrated superior performance compared to individual LSTM and other decomposition-based models (EMD-LSTM, EEMD-LSTM, CEEMDAN-LSTM).
• Significant reductions in Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) were observed across maize, palm oil, and soybean oil price data.
• Statistical tests (TOPSIS, Diebold-Mariano) confirmed the enhanced prediction accuracy of the VMD-LSTM model.

Conclusions:

• The proposed GA-optimized VMD-LSTM model offers a robust and accurate solution for agricultural commodity price forecasting.
• This advanced forecasting tool can significantly aid stakeholders in making more informed economic decisions.
• The model's ability to capture complex price dynamics represents a substantial advancement in agricultural market analysis.