Enhancing agricultural commodity price forecasting with deep learning.

Area of Science:

• Agricultural Economics
• Data Science
• Time Series Forecasting

Background:

• Accurate agricultural commodity price forecasting is crucial for market planning and policy in agriculture-dependent economies.
• Price volatility, influenced by weather and market demand, presents significant challenges for traditional forecasting methods.
• A comprehensive evaluation of various forecasting models is needed to address these challenges.

Purpose of the Study:

• To evaluate and compare the performance of traditional stochastic, machine learning, and deep learning models for agricultural commodity price forecasting.
• To identify the most effective models for capturing complex temporal patterns and volatility in commodity prices.
• To provide insights for improving market interventions, crop planning, and risk management strategies.

Main Methods:

• Utilized daily wholesale price data for 23 commodities from January 2010 to June 2024.
• Assessed traditional models (ARIMA), machine learning (SVR, XGBoost), and deep learning approaches (MLP, RNN, LSTM, GRU, ESN).
• Compared model performance using error metrics such as Root Mean Squared Error (RMSE) and Mean Absolute Percentage Error (MAPE).

Main Results:

• Deep learning models, specifically Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU), demonstrated superior forecasting accuracy.
• GRU achieved significantly lower RMSE and MAPE for commodities like onions and tomatoes compared to the ARIMA model.
• Deep learning models effectively captured complex temporal patterns and nonlinear dynamics inherent in agricultural commodity prices.

Conclusions:

• Deep learning techniques, particularly LSTM and GRU, offer a more reliable approach to forecasting volatile agricultural commodity prices.
• These findings support enhanced market interventions, better crop planning, and more effective risk management for stakeholders.
• Future research should explore hybrid models and incorporate external data, such as weather information, to further boost forecasting accuracy.