Agricultural Commodity Price Prediction using Long Short-Term Memory (LSTM) based Neural Networks

Abstract

Background: Agricultural price forecasting is one of the research hotspots in time series forecasting due to its unique characteristics. In this paper, we develop a standard long short-term memory (LSTM) for accurately predicting a nonstationary and nonlinear agricultural price series. Methods: An LSTM model effectively analyses and captures short-term and long-term temporal patterns of a complex time series due to its recurrent neural architecture and the memory function used in the hidden nodes. Result: The empirical results using the international monthly price series of maize demonstrate the superiority of the developed LSTM model over other models in terms of various forecasting evaluation criteria. Overall, LSTM model shows great potential for improving the accuracy and reliability of agricultural price predictions, benefiting farmers, traders, and policymakers in making informed decisions.