Navigating soybean price volatility: A deep learning perspective

Abstract

Soybean, a significant oilseed crop, has become increasingly vital in India over the past decade, serving as an essential protein source for both human consumption and livestock feed. With soaring production and demand, especially in regions such as Madhya Pradesh, Maharashtra, Rajasthan, Karnataka, and Gujarat, there's an amplified need for reliable soybean futures price predictions. Forecasting in the futures market is not only of immense value but also technically challenging. This study delves into a comparative evaluation of soybean futures prices using various deep learning models, including Time Delay Neural Network (TDNN), Recurrent Neural Network (RNN), Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM). Our findings reveal that the LSTM and GRU models substantially outperform the TDNN and RNN in terms of forecasting accuracy. Specifically, the LSTM model emerges as the pinnacle, delivering unparalleled directional forecasting results. The efficacy of the models was further assessed using Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE), wherein LSTM was identified as the most representative model for soybean price predictions. This research provides pivotal insights for futures price forecasting applications, presenting a robust model that could serve as a crucial policy tool for farmers, processors, and traders.